Validation: how to validate accuracy

[morse]

Dear all,

What is your opinion on this issue?

ICHQ2(R1) states that accuracy "should be assessed on samples spiked with known amounts of impurities".
"In case where it is impossible to obtain samples of certain impurities, it is considered acceptable to compare results obtained by an independent procedure".

I have to validate an analytical method. I can't access one of the related impurities to do the spiking recovery test. I can't apply another procedure either as there isn't any that is already validated.

Do you have any idea on how I can execute this test (accuracy)?
Hopefully you can give me a hand.

Thanks a lot,

Morse.

[TimB]

Do you have a reference standard?

What about using Standard addition? A sample is assayed, a known amount of pure active constituent is added, and the sample is assayed again. The difference between the results of the two assays is compared with the expected answer.

[morse]

Dear Tim,

This is th problem.
I don't have this impurity nor any reference standard of it..

Thank you for your kind answer.

[tom jupille]

Read the rules like a bureaucrat. It says "by an independent procedure", not necessarily "by a validated independent procedure".

Is there an element of sophistry there? Yes, but if it works . . .

[morse]

Thanks Tom!

You are right, actually.

But, again, I don't have a standard.
Do you think I would be allowed ti quantitate in % Area?
How can I do the linearity test?

Thanks a lot for your help.

M

[shaun78]

Everything that you try to do will require you to have an actual standard at some point in time.

Your best option, especially if regulated by the FDA or similar agency, is going to be take an aliquot of your material and get it qualified as a standard.

Sure, you have the chicken before the egg problem because the first time you are going to be qualifying the material against itself ... but after that you will be golden.

On a side note ... In specific instances, one is also allowed to infer accuracy from linearity. I have actually done this numerous times in the past and have had the FDA crawling all over the work. They (FDA) had no problems with it.

[aceto_81]

On a side note ... In specific instances, one is also allowed to infer accuracy from linearity. I have actually done this numerous times in the past and have had the FDA crawling all over the work. They (FDA) had no problems with it.

Can you give an example workflow of infering accuracy from linearity?
Just to get an idea how things can be done.

Ace

[Sallybeetle]

I assume you are referring to ICH Q2b.

I have a similar situation. I am trying to validate an assay method as an impurities method.

In the case of no known impurities or unobtainable impurities, can accuracy, linearity, etc. of impurities be inferred by performing validation using the assay main analyte down at the impurities level, e.g. reporting and identification levels (0.3% and 1%)?

[morse]

dear All,

Thanks a lot for taking part in the discussion.

Schaun,
How can I succeed in qualifying an aliquot of this impurity (if I ever manage to purchase it)?
Do I need to have some of the tests performed by an external qualified lab?
I reckon an absolute technique assay needs being necessarily done (like NMR, DSC) and also elementary analysis if I am not mistaken.

Sally,
as for linerity and accuracy.. I assume you can't perform them without the actual compounds. You can carry out linearity of the main ingredient down to a a very low concentration but how can you prove your method is linear with respect to the impurity in this way?

[Sallybeetle]

A few years ago, this was the approach we used when a compound was new / just off patent, and we could not find literature or compendial references to known impurities; or, we just plain could not obtain a specific impurity.

I guess the assertation was that we were proving the method to be accurate and linear for at least the main compound down at the 0.3% reporting and 1% identfication impurities levels. Since no known impurities were available, that was the next best option. When impurities became available, we would validate the method for those compounds at that time. We assumed an impurities response factor of 1 for the unknown / unobtainable impurities compared to the main compound.

Are you calculating the impurity content against external API reference standard, peak area percent of main compound in same injection, or specific impurity reference standard?

[shaun78]

Accuracy can be inferred from linearity in instances where significant amounts of standard are not readily available or the cost of obtaining the standards is very high. Accuracy may be inferred from linearity assuming that that the intercept is clost to zero (few percent) and the r-squared is 0.999 or greater.

The testing to qualify a standard does not have to be done by an outside lab, but most labs do not have everything that is needed to get the testing done. Therefore, it gets contracted out a lot. You are correct on the testing that needs to be done: chromatographic assay of some sort, IR, NMR, and DSC are generally accepted as the minimum.

I may be able to help you out a little more as I believe I have misred some of your post. So, you are working on an impurities method for some sort of an API, correct? The impurities produced in the manufacturing of the API are not available to you can can not be easily produced, correct?

If the above two are true, then you are allowed to validate an impurity assay based off trace analysis of your active as Sallybeetle suggested. For example, validate accuracy by spiking 0.1% to your impurity limit of active. The justification is that you can't get impurity standards and the imurities are somewhat related to the finished product so it can be reasonably argued that the impurities are going to behave the same way.

I too have done validations based off the above in the past, the FDA has looked at them and did not have any comments about why I choose to validate an impurity assay based of the active (I stated why in the report).

[morse]

Dear Sally and Schaun,

Thank you for your advice. Very much appreciated.

Schaun, I think your reasoning about accuracy through linearity approach is very sound, though sometimes I expect the inspectors to be somewhat dull (not all, of course). This means they might a bit "legalistic", sometimes. Where did you get the information you've given from? I do agree with you, though I would like to know the source to support this choice in case it's needed.
I read the guideline and it appears to me (I might be mistaken-haven't checked right now) you're allowed to validate accuracy through linearity only when it comes to an assay method, not related impurities.


As for Sally, I, again, agree with your approach. What else can you do? It's, as you said, the next best. Though, it's not written on any guideline. This is my concern and worry.
Also beacause, as I said before, you would be validating the linearity of your API at very low concentrations and not the impurity at issue linearity.
I happen to work with methods made up in that way. Yet, it never happened to me to be lacking specified impurities when validating.

Yes, Schaun.. you understood rightly. The two considerations you've written describe my situation.

Thank you again.
It's great to talk to you.

Morse

[shaun78]

Take a look at USP <1225> under the accuracy section of the analytical performance characteristics.

Assessment of accuracy can be accomplished in a variety of ways, including evaluating the recovery of analyte (percent recovery) across teh range of the assay, or evaluating the linearity of the relationship between estimated and actual concentrations ...

I don't think the USP is specifically talking about analytical assays in this section.

Furthermore, I worked for a large pharmaceutical company in the past where every single method I worked on inferred method accuracy in this manner. Becuase of regulatory issues we had, the FDA looked through every single validation we did at the time. We made proprietary compounds where impurity standards were absolutely not available.

[HW Mueller]

Is this a nightmare or am I understanding this correctly: The authorities have given the OK for validating the accuracy of an analysis even though it was not known what the analyte´s detector response is?

[lmh]

Absolutely, HW Mueller. I'm glad someone has brought this up. I thought I was just being stupid and not understanding properly.

I can't see how accuracy can be "inferred" from even the most beautifully straight and repeatable calibration curve (in general, for a typical detector). If you don't know the detector response, you have no idea how much material is present in the sample. All you know is that it is X times the amount in another sample. This is Precision, not Accuracy. Taking the USP <1225> quote, when you evaluate the relationship between estimated and actual concentrations, the relationship should not only be linear, but be 1:1, and the actual concentrations must be actual. Unfortunately that leaves no way to avoid using a genuine standard.

Of course if you haven't got a standard, you're stuffed, but it's better to be stuffed and admit you don't know something than to pretend the data are better than they are (in my view).