FDA - Integration Algorithm Consistency

[mactavish606]

Our HPLC data isn't audited by the FDA, but I would like to get into the habit of doing everything possible "by the book." So, does the FDA prefer to see the same algorithm used for all samples run of the same analyte regardless of the time they were run? For example: should all samples tested in order to determine the amount of caffeine in them on January 1, 2011 use the same algorithm as those samples tested for the caffeine content run on June 1, 2011?

I would think that as a column ages the integration would change accordingly in order to properly integrate the peaks, but it could be argued that if the column ages to the point that you have to change the algorithm then you need to replace the column.

When I say "integration algorithm" what I am referring to is - on our Perkin Elmer HPLC (Chromera) system at least- is that integration is determined by 3 numbers: bunching factor, area threshold, and noise threshold (as this is the only HPLC software system that I have operated I have no idea if this is consistent across all machines - once again, my guess would be that it's pretty similar but with the slight variations that come with various companies but since I have virtually zero experience I'll refrain from making assumptions).

Thanks for all the help!

[bisnettrj2]

I don't have experience with FDA requirements, but I can say this: As the column ages, you will need to recalibrate to account for deteriorating plate heights and peak shapes (as long as you are within the requirements set by the FDA). When you re-calibrate, you could probably reset the integration parameters to account for the differing response from the last calibration. However, you must keep those parameters for the entirety of that calibration curve, to make sure the integrations for samples are based on the same algorithms as the calibration curve they are based on.

I do know that auditors do not like manual integrations. Messing with the integration parameters after an initial calibration seems like an end-around to performing manual integrations, to make up for inconsistencies or deterioration in the separation. Just my opinion.

[DSP007]

Our HPLC data isn't audited by the FDA, but I would like to get into the habit of doing everything possible "by the book." So, does the FDA prefer to see the same algorithm used for all samples run of the same analyte regardless of the time they were run? For example: should all samples tested in order to determine the amount of caffeine in them on January 1, 2011 use the same algorithm as those samples tested for the caffeine content run on June 1, 2011?

Yes of course .
If on 1 Jan 2011 you was used coffein solution in water but in Jule you plan use naphalene solution in benzene - this "algoritm" should generate many auditor question
Needless to say, that should be used the same dilutions, the same chromatographic conditions and the same method of data processing. To be able to evaluate changes in the work of the unit mathematically.

I would think that as a column ages the integration would change accordingly in order to properly integrate the peaks, but it could be argued that if the column ages to the point that you have to change the algorithm then you need to replace the column.

When I say "integration algorithm" what I am referring to is - on our Perkin Elmer HPLC (Chromera) system at least- is that integration is determined by 3 numbers: bunching factor, area threshold, and noise threshold (as this is the only HPLC software system that I have operated I have no idea if this is consistent across all machines - once again, my guess would be that it's pretty similar but with the slight variations that come with various companies but since I have virtually zero experience I'll refrain from making assumptions).

Thanks for all the help!

Correct start and end points of integration. So your peak counted correctly. After all, you do not interisuet status column, and change interisuet statistical convergence, linearity, and sensitivity and noise of the device itself.Agilent in IQ/PV controlled device without column, just to the capillary.

[unmgvar]

mactavish606 as DSP007 pointed out ( but I feel too little)
FDA regualtions are in place in order to make sure you and everybody else can trust your resuslts and that no foul play is done, wether with or without purpose (let's not forget that it can cost and it did in the past, someones life)

so you need to think in a very simple manner, applicating this is sometimes a lot more harder to do :
all your work needs to be performed and documented in such a way that you can show consistent reliable and accurate repeatable work that is most importantly capable of picking up what goes wrong.
the major approach for this, which makes things hard is simply to go on the bases that you 100% know what you have at every single step of the way in your procedure.

one major and important step for everything that you do is to set a frame of parameters and boundaries that you can check
for example:
you talk of the column behavior,
yes retention time of your compouds will change over time as the column ages, but in no way is there a method even for a single peak where you let it move from 15 minutes to 5 minutes and still say things are ok. even if you inject a standard as well.
for the algorithem of softwares, if you are talking of a nice looking with strong signal peak then there is no reason to change the type of integration settings,
but in the case of a complex chromatography this is close to impossible and the goal is to use the avalaible tools in order to report correctly displayed integrated peaks.

you said do things by the book, and a book it is that you need to writte for yourself.
you need to implement SOPs and protocol and work by them (your book). each work should be filed (another book) and archived for historical reference, library now(this is why paperless is so strong in pharma these days)
if you can always say safely for every step (i cannot start to explain that this includes all steps, direct and indirect) the 5 Ws
Who did it
what was done
when it was done
why it was done
how it was done

within the guidelines that you have set for yourself

then you are cGMP/cGLP
so that when things go wrong you can play "Sherlock Holmes", and not guess, but show and prove what was wrong and how you fixed it