I have a question regarding linearity in the validation of a HPLC assay method: Apart from the study of residuals - ANOVA (outliners) and calculation of the regression curve (reporting the correlation coefficient) are there any critaria for the acceptance of the curve regarding intercept (standard error on intercept or correlation between Area100% and Area0% (intercept)?

The question because i found a SOP in which is stated that Area(100%)/Area(intercept)*100 should be < 2%. I don't know from where it comes!!

Yes, there are 2 additional criterias to be satisfied in a linearity curve.
1. Bias should be <2%
Bias=(Y-intercept/average response at the assay level)X100

2. Response factors at each data point should be within +or-10% of the Average Response Factor.

These criteria come from a course that I attended and do not know who originally put forward. I know people in pharmaceutical companies (myself) follow these guidelines.
I personbally found, for biomolecules it is not possible to satisfy all these criteria
Hope this helps.

Ananda S.

RP,

I think you got your numerator and denominator switched. Like Ananda, I prefer a response factor approach. see http://www.lcresources.com/discus/messages/3282/3214.html?WednesdayFebruary2620031201pm

Good Luck.

Thank you for your response!!!

Ananda states; "Response factors at each data point should be within +or-10% of the Average Response Factor."

If you make a plot of this and translate into a x-y graf you will not see "a linear curve" suitable for calibration in e.g., shelf-life studies or product control by HPLC. It is indeed likely that the intercept will be within a 95% confidence interval.

I would expect that most analytical HPLC methods for such purposes have a response factor <± 2%? For biomolecules it might be different.