Long-term HPLC RRF Stability.

[Sally_F]

This should be easy: I'm looking for references describing the long term stability of RRFs in HPLC methods.


Complete Info:
My lab's used a RRF method on our HPLC units for the last 7 years. Every analysis is preceded by a standard that we use to re-compute the RRF for the instrument. We use this pre-run RRF to prove statistical control over the instruments. Problem is that minor differences in equipment set up, standard preparation, etc. frequently give us control errors for our on-side of aim rule. (Frequently as in 5 times a year, or so). I've reams of data proving the stability of the method, and proof that the minor fluctuations we see in the RRF do not significantly affect the methods' uncertainty.

So I'm trying to eliminate the on-side of aim control rule for this instrument, but can't because I can't provide any external sources touting the stability of RRF methods in general. There are plenty of method-specific sources, but nothing general enough to take to my superiors as proof.

Help please?

ps. This is not a pharmaceutical application.

[Sally_F]

So nothing? There's nothing out there around this?

[Dan]

Sally,

I am not aware of any such references for RRF in the literature, regulatory guidances or in any internal documentation in the places I have worked.

Mostly, I have seen that the RRF values are determined once during the method validation. The idea is to get suffcient measurements to ensure a good average value (but not take too much time doing it as you are on a deadline).

Two rules that I suggest to apply:

1) The RRF value should be determined to only 1 decimal for values greater than or equal to 1 and to 2 decimals for values below 1.

2) If you are using a control chart for the long-term stability of the value, then be sure to use a practical significance and not necessarily a statistical significance to determine or evaluate the variation. Finding a difference of 5% in an RRF value over time may be statistically significant, but is it of a practical significance?


Take an example for the impurity analysis of a drug product:

The initial RRF is 0.72 and the new RRF is 0.76. That's a difference of about 5%. Now, suppose the quantitation of the impurity gives a result of 0.21 % w/w when an RRF of 0.72 is used in the calculation. Re-calculating using an RRF vlaue of 0.76 gives a result of 0.21 or 0.22 % w/w, depending on how the rounding went during the calculation. So, a 5% difference between the old and the new RRF values may not even change the result. (You do have to be careful when the impurity result is close to the specification limit.)

In method development and validation work, I have often found it better to apply practical significance rather than statistical significance when evaluating data.

Regards,
Dan