By Nick Pittman on Tuesday, July 6, 2004 - 03:42 am:

Hi there,

I've been set the (enviable) task of looking at the type of integration to use in the event of a small impurity (usually NMT 1%) appearing on the back of a tailing drug substance main band peak. Usually, we would drop a line down to the projected baseline, thus overestimating the impurity area. However this has recently led to problems so we may be reassessing this approach.

I understand previous threads on a similar subject have mentioned the idea of quantification using peak height, however I am unsure how well this would be received in a GMP environment. Unfortunately, tweaking chromatographic methods to provide better separation is not an option due to the required turnaround of samples in our group.

Any advice people could provide, or any references to use as a starting point would be extremely gratefully received!

Thanks all!

Nick

-------------------------------------------------------------------------------------------------------
By Anonymous on Tuesday, July 6, 2004 - 01:46 pm:

sometime the "tangent skim" integrationparameter works

-------------------------------------------------------------------------------------------------------
By Jimmy on Tuesday, July 6, 2004 - 05:00 pm:

I agree. Generally, if you have a very large difference in the size of the unresolved peaks, then skimming the small one "off" of the large one should work fairly well. When the two peaks are about the same size, using dropline integration works well.

It's in the in-between cases where it is becomes impossible to integrate in such a way as to get accurate areas or heights.

-------------------------------------------------------------------------------------------------------
By B.Buglio on Tuesday, July 6, 2004 - 07:10 pm:

A lot of misery both now and in the future will
be avoided if you,in fact, do the right thing:
spend some time to optimize the separation. A good
approach would be to use the opposite separation
mode (if your in RP go to NP) to elute the small
peak before the large one.

-------------------------------------------------------------------------------------------------------
By rob burgess on Wednesday, July 7, 2004 - 07:39 am:

Nick,
You say you've encountered problems with perpendicular drops. What problems specifically?

You could always do both and compare for all samples you are looking at. Maybe the trends for the impurity is the same for both integration methods.

There is a book specifically on Integration methods (by N. Dyson I believe) or you could always consult the HPLC method devlopment bible. Maybe these texts could throw up some more specific guidance.

I know many guidelines incorporate taking into account tailing factors, but as per a previous post on resolution, its not much use if your mneasuring a peak at less than 1% level whereas Tf is measured at 5%.

In the end for impurities in pharmaceuticals I would have thought that overestimating was better than underestimating, just to be on the safe side.

-------------------------------------------------------------------------------------------------------
By tom_mizukami on Wednesday, July 7, 2004 - 12:04 pm:

Hi Nick,

I would be leery about changing the integration method for a drug substance method in a GMP lab. Was this method validated, was impurity accuracy and linearity examined, what integration method was used during validation?

If you still have all of the data files around from validation you may be able to reprocess them to examine the effect of a change in integration method. Many data system have the ability to perform exponential skims or tangent skims. I would start with the user manuals. Most vendors consider their algorithms proprietary and are poorly documented. Tracing back to a source reference may be difficult. Good luck.

-------------------------------------------------------------------------------------------------------
By rob burgess on Thursday, July 8, 2004 - 01:45 am:

Surely the best method to use for integration is manually. Most of the algorithims I know for integration are never as accuarate as the human eye.

-------------------------------------------------------------------------------------------------------
By Nick Pittman on Thursday, July 8, 2004 - 09:02 am:

Thanks for your replies everyone!

The problem we have seen recently is that an impurity in a batch of material passes specification when a tangiential skim is used, but fails when a drop line is employed.

I agree that more work is needed on the method, but unfortunately i am not in a position to do this. The methods we see problems with are generally developed by another group and are sometimes in quite early stages of development. They are validated methods, but the level of validation will depend on the usage of the data generated.

I'll have a look for the book you mention Rob - many thanks.

Any more advice would be great - thanks for all the help so far!

Nick

-------------------------------------------------------------------------------------------------------
By B.Buglio on Friday, July 9, 2004 - 04:44 pm:

This is a troubling thread. As reported, the
method in question was validated and transferred
to the end user. In accordance w GMP procedure a
method transfer study (as part of the tech
transfer) should have been run. At that time
retained samples would have been run by the
developing and end user labs. Once the parameters
for obtaining the same assay values(in both labs)
were established they should not have to be
changed. If these parameters don’t work something
has changed in the method e.g. resolution, tailing
factor etc.

Changing the integration method borders on
subterfuge and you should not be doing this. Your
supervisor should be in contact w the developing
lab and investigating the problem.