Accuracy through Linearity Validation

[morse]

Dear All,

I read a topic on the issue above (it's locked now) and I'd like to ask your opinion on a few concepts.

-Is there article on publication on the possibility ofv validationg accuracy through linearity?

-How would you practically do it? I know one would back calculate the analyte concentration from the regression line. What will be the right acceptance criteria? Would you increase it going toward lower concentrations?

-Last but not least. It seems from ICH that this criterion can be applied only to assays and not to impurities. Am I wrong?

Thanks a lot for your help,
Morse.

[krickos]

Hi

Would be intresting to hear how this concept has been used.

As for ICH (ICHQ2R1) I assume you refer to bullet c) below?:

4. ACCURACY
Accuracy should be established across the specified range of the analytical procedure.
4.1. Assay
4.1.1 Drug Substance
Several methods of determining accuracy are available:
a) application of an analytical procedure to an analyte of known purity (e.g. reference material);
b) comparison of the results of the proposed analytical procedure with those of a second well-characterized procedure, the accuracy of which is stated and/or defined (independent procedure, see 1.2.);
c) accuracy may be inferred once precision, linearity and specificity have been established.

The C option is not viable to my knowledge when it comes to quantification of impurities, spiked solutions should be used.
But if someone has used it I would like to know

[morse]

Thanks for answering, Krikos.

Yes, I refer to point C.
Why do you think this approach is viable for assays and not for impurities?

I'd love to see this way of validating accuracy used, too.

Morse

[krickos]

The potential pitfall I see is this:

In assays. Here you tend to use the "same way" of evaluation as in the intended final method when you do linearity and accuracy.

In impurities you do a linearity investigation of minimum 5 points (think external calibration curve with minimum 5 points) but when you do accuracy, you may use internal normalisation (area %), weight/weight %...
You will likely not use a 5 point calibration curve when you run the method routinely or?.
Hence the result evalution does not match between validation and actual method.

Sure you can study linearty for impurities by spiking impurities at different levels into solutions derived from a stock sample solution and then evaluate linearity and accuracy at the same time but thats not what you asked for.

Or I am missing something?

[HW Mueller]

Doesn´t the word "drug substance" include its impurities? Doesn´t the word "assay" refer to analyzing the whole pill, or whatever is to be analyzed?

[morse]

Krikos and HW,

Thank you a lot for your interest in this topic.

Krikos,

I think there's something I am not getting of what you're saying.Could you please try to explain it again? As far as our company is concerned, we don't execute linearity while performing an assay. We routinely use the external standard quantitation and assume from the method validation that the method response is linear in the range we are working.
We actually operate in the same way with the sample related impurities. We generally use an external standard quantitation.
So, I still can make out why ICH allows to infer the accuracy from linearity in assay test but not for impurities.
In our validation procedures, we actually operate in the same way.. at least five levels of concentrations both for assay and for impurity determination.

HW,

I believe you think that by saying drug product or drug substance they mean the whole product, impurities included. And by assay they mean assay of the API or assay of the impurities (determination).
I agree with you but can't understand why ICH Q2B add a further paragraph called "impurities" (4.2) to specifically deal with realated impurities. You know what I mean?


thanks a lot to both of you.

Morse

[HW Mueller]

No, it seems to me that they use the word "assay" as I learned it, namely, that it is a synonym for analysis. If true this would mean they refer to the analysis of everything in the sample, then detail in the subheading the anlysis of the "drug substance", which I see as a synonym for the main drug + impurities from it. I would expect that the assay/analysis of other parts of the sample are detailed elsewhere.
I have immense trouble with other interpretations since they seem to go in the direction of guesswork.

[morse]

HW,

Yep, it was what I meant. Assay as analysis, determination.
I wasn't very clear.. sorry about that, but I know what you meant.

Yet, it is not very straightforward.
I haven't found any article or publication yet.

Thanks for clarifying.
Morse

[Alex Buske]

Pharmacopoeial method usually have, amongst others, a assay and a related substances section. The assay only deals with API, while the rs method covers its impurities. Sometimes this can be done in one test, often it is easier to split the two methods.

Alex.

[Peter Apps]

"Assay" in the world of drugs probably goes back to the dark days before active ingredients could be specifically analysed, or even identified. Batches of "drug" in the old fashioned sense of whole material were assayed by various semi-empirical tests that showed how strong their effects were.

A similar approach was followed in the early days of vitamins - the one that sticks on my mind is the rat assay for vitamin D - feed a bunch of young rats on a vitamin D deficient diet spiked with the material whose vitamin contnet you wanted to know, and after a few weeks kill them, dissect out their femurs and measure how curved they were.

I think that the mouse bioassay is still the official method for some marine shellfish toxins.

Peter

[shahis77]

IF YOU ARE TALKING IN BIOANALYTICAL WAY THEN FOR LLOQ THE LIMIT IS 20% AND FOR ALL OTHER CONCENTRATIONS IT IS 15%..

[JGK]

IF YOU ARE TALKING IN BIOANALYTICAL WAY THEN FOR LLOQ THE LIMIT IS 20% AND FOR ALL OTHER CONCENTRATIONS IT IS 15%..

there' no need to shout

[danko]

Who’s talking about limits anyway? Or have I missed something?

Best Regards

[JGK]

For non-bioanalytical assays there appear to be no fixed acceptance criteria. In the years I've been working within the regulatory environment I ran two types of assay:

Calibrating between ±50% around a fixed point. with this "narrow" calibration range we validated using 5 calibration points and had accuracy acceptance criteria of 100±2% at each point.

Calibrating across a wider range (1-2 orders of magnitude). with this type calibration range we validated using 6 (non-zero) calibration points and had accuracy acceptance criteria of 100±5% at each point and 100±10% at the LLOQ.

[alicee!]

The potential pitfall I see is this:

In assays. Here you tend to use the "same way" of evaluation as in the intended final method when you do linearity and accuracy.

In impurities you do a linearity investigation of minimum 5 points (think external calibration curve with minimum 5 points) but when you do accuracy, you may use internal normalisation (area %), weight/weight %...
You will likely not use a 5 point calibration curve when you run the method routinely or?.
Hence the result evalution does not match between validation and actual method.

Sure you can study linearty for impurities by spiking impurities at different levels into solutions derived from a stock sample solution and then evaluate linearity and accuracy at the same time but thats not what you asked for.

Or I am missing something?

i have a few doubts very close to wat ur talking abt..

i dont have purified impurities to spike.. so i dunno how to prove accuracy using this approach.
how i plan to do it is using my purified protein.. assuming say an oxidized impurity will not behave very differnt from the native purified form that i m using to establish accuracy..
i plan to do a linearity and backcaluculated graph based recovery..
i want to know.. u mentioned 5 pionts.. wat is the range that you would evaluate this for.. near the impurity range.. i.e if my injection amount was 15 mcg.. do i do it near 0.2-2% of this.. where my impurities would lie.. or 0.2-120% of 15 mcg.. i.e. full range..
is a full range linearity expected?

also..
for quantitation of impurities.. i could do a 5 point curve between 0.2-2% or do an area% like u said.. i get a difference in these values.. as i donot expect full range linearity.. is this unusuall to happen.. what is the preferred approach to address this..

thanks!