GPC

[Luci]

Does anyone know how to perform the linearity characteristic in the method validation for GPC?

Thanks
Luci

[tom jupille]

For quantitation or mw determination? If quantitation, it's the same as any other type of LC.

[Luci]

It is Mw determination. Standards curve have 6 polystyrene standards. Can I run a curve as normal and calculate the correlation coefficient with third degree fit? or do I have to do spikes? If I have to do spike, how?

[AA]

Mw determinations are based on elution volume (often expressed as retention time) and those curves not really linear.

[tom jupille]

those curves not really linear

Which is why she's using a cubic (3rd degree polynomial) fit. I've run into a lot of people (particularly regulatory people) who mis-use the word "linearity" to be a synonym for "predictability". I had one person tell me he had gotten good linearity with his quadratic fit!

But back to the original question. I'm not sure what doing "spikes" would do for you in a mw determination. What you need to do is demonstrate that your fitting function is adequate to your data. In addition to the usual parameters (intercept and coefficients for x, x^2, & x^3, you will need to evaluate the R^2 and look at the residuals. You can also run a p-test to see if the cubic fit is appropriate. All of those can be folded into an LOF ("lack of fit") test.

I'm not a statistician, but a good place to start is with the Coleman and Vanetta articles in American Laboratory, specifically sections 22 and 23:
http://www.americanlaboratory.com/913-T ... t-Details/
http://www.americanlaboratory.com/913-T ... -Fit-Test/

[Luci]

Thank you Tom Jupille.

Luci