HPLC peak integration method

[S.M.PANDYA]

Dear Experts

One laboratory is using valley to valley integration method for peak area calculation in Related Substances test performed by HPLC & another Lab is using Auto Integration method for the same .

Now problem is by Valley to Valley integration method result of one impurity is about 0.07 % & the same impurity if calculated by Auto integration method it shows 0.12 % impurity . Due to this 0.12 % imp. is considered as failed batch .

So could you please advise me in this situation which method of integration is to be followed as per law of HPLC technique .

[Bruce Hamilton]

Dear Experts

One laboratory is using valley to valley integration method for peak area calculation in Related Substances test performed by HPLC & another Lab is using Auto Integration method for the same .

Now problem is by Valley to Valley integration method result of one impurity is about 0.07 % & the same impurity if calculated by Auto integration method it shows 0.12 % impurity . Due to this 0.12 % imp. is considered as failed batch .

So could you please advise me in this situation which method of integration is to be followed as per law of HPLC technique .

Without seeing the chromatograms or details of the methods, the law requires the methods to be validated. If they have both been validated, you have to conduct experiments to identify with is correct for the sample, and suspend the flawed method.

The obvious suspicion is that the v-v method doesn't reach at baseline, but it could also be that the auto integration method is using v-v but with different threshold setting.

Bruce Hamilton

[Uwe Neue]

In cases of marginal resolution, different methods of peak identification and allocation of areas to each peak will give you different results. In your case, you need to specify the method of integration.

To explain (and I think this is close to your case): you have a giant parent peak, and an impurity peak that elutes on the tail of the parent peak. If you simply decide to drop a vertical line to the baseline in the valley between both peaks, the integration of your impurity peak will include a portion of the parent peak. You can subtract the area of the parent peak to obtain a better peak area for the impurity peak.

But you MUST specify the correct way of doing this in the method, i.e. BEFORE doing the analysis.

[grreddy]

Dear Pandya,

Always you have to integrate Base to Base which is advisable against vally to vally because some times two impurities will elute closely. In this context you have to check the validation data where how they integrated.
In case of Water's software you can perform both auto integration as well as Manual integration by setting up threshold level where as in Agilent chem station and Class VP it is manual integration and again it is upto the laboratory that which method they use to follow.

In our case in waters software we are using auto integartion and for chem station we are using manual integration for impurities against method validation report along with justification for manual integartion.

G.R.Reddy,
AR&D,
BIOCON
India

[Mark Tracy]

There were a pair of articles in LCGC not too long ago that analyzed the performance of various integration methods using known standards. Valley-to-valley consistently underestimated the true values. Bruce's advice is right, and I would bet that v-v will be the loser when you do the experiments.

[JM]

There is NO universal right integration method (including auto-integration). You have to look case-to-case basis. It all depends on type of peak you are integrating V-to-V may be Ok for resolved peaks but not for semi-resolved peaks , there are several type of integration based on peak type and shape and most of today’s software have em all. It is the discrition of analyst to chose the right one.


JM

[Bruce Hamilton]

There is NO universal right integration method (including auto-integration).
JM

I'm waiting for a peak to appear, so I shouldn't write this, it will come back to haunt me....

The universal right integration method involves...
One chart recorder producing on-scale peaks, a chromatogram on chartpaper, one pair of sharp scissors, one analytical balance, one sober and sighted operator....

If you have never used cut and weigh, you don't know what you're missing

Bruce Hamilton

[JM]

Hi Bruce,
I have used cut and weigh method way back in 90's , even this is also not problem free. Paper thickness issues and you need to decide where to cut. For assay of active, integration does not make much difference but for semi-resolved impurities it does.

There is very good article in LCGC on this topic in 2 parts,

http://www.lcgcmag.com/lcgc/article/art ... 9&pageID=1

JM

[HW Mueller]

Well, this has been mentioned several times before: A way to partially get around this is to devise two analytical methods. I also said this before: There is still some pride here that cortisol could be detected within a range of about 10% with two different 3 step chromatographic (mostly) methods. That was incomparably better than RIA or ELISA (immuno methods).

On cutting and weighing: What memories...the ones who could hold a scissors properly had quite an advantage. (We found that strip chart paper was surprisingly uniform in thickness).

[Peter Apps]

How sure are you that the difference in results is due ONLY to the diference in integration ? Have you exchanged a set of raw signal data between the labs and had them try thier integration son the other lab's data ?

Assuming that everything else is under control you can avoid this problem by optimising the chomatography to give resolved peaks, or by calibrating with matrix-matched standards with the same integration parameters for standards and samples of course.

Could you post chromatograms that show where the integrator is drawing the baseline for each method ?

Peter

[DR]

[anupama]

Usually it's from baseline.Could you post your chromatogram

[Dan]

S.M.Pandya,

I think that Peter Apps has a good recommendation. Did you check for integration differences for the acquisition parameters? To follow up on it, I suggest that you check the data acquisition frequencies of the two sites. You should be collecting enough data points per peak to get proper integration. You should have at least 20 points per peak (30 or more points per peak may be better). I have seen bad integration resulting from having too few data points, especially for small peaks.

Lastly, be careful about trying to change the integration to change the 0.12% to a passing result. NO auditor would like that to happen. Having bad integration acquisition parameters can provide for a good reason for the change. However, it may mean re-aqcuiring new data on the samples.

Regards,
Dan

[Russ]

The talk of cut and weigh brings back some memories. Anyone else use the "high tech" alternative of a planimeter? For those who are wondering what I am talking about:

http://persweb.wabash.edu/facstaff/foot ... IMETER.HTM

[rolpe]

Hi,

the integration of not properly resolved peaks is always difficult. No matter how you integrate, it's never 100% correct. So I would recommend to increase the selectivity of your method (increase flow, decrease column temperature).
If this is not possible I would suggest to use the method best for you. I know this is not the analytical correct way but if the difference is only 0.05 % this is in the analysts freedom. I would integrate by my "stomach feeling".
But best would really be to get the Peaks properly resolved.

Best Regards
Rolpe