Standard Curve Linearity

[Sallybeetle]

I have always seen a specification on standard curve correlation coefficient as greater than 0.98 for general GMP regulated analyses in a QC laboratory. However, recently our lab received a comment that the correlation coefficient specification must be set to a minimum of 0.99.

Does anyone on the board know of a reference on this subject?

[danko]

Actually it is the same requirement. It’s just that the 0.99 is the correlation coefficient, whereas the 0.98 value is the coefficient of determination, which is calculated as follows: 0.99^2 = 0.98.

Best Regards

[krickos]

Hi

Well now I am a bit confused. FDA (CDER Reviewer Guidance fron Nov-1994) states that r should not be less than 0,999 in most circumstances (r-square not less than 0,998).

Also have a vauge memory (unsure if it from an inspection or formal FDA comment) that FDA at some point stated a r of 0,999 was pretty much expected for HPLC applications.

[danko]

Hi Kickos,
You’re absolutely right about the FDA recommendations/expectations.
My comment/explanation, however, was authority neutral (i.e. regardless of who expects what). The point is that some people forget to specify the linearity definition when specifying the requirements.
In your example “râ€

[krickos]

[quote="danko"]Hi Kickos,
You’re absolutely right about the FDA recommendations/expectations.
My comment/explanation, however, was authority neutral (i.e. regardless of who expects what). The point is that some people forget to specify the linearity definition when specifying the requirements.
In your example “râ€