Resolution, method of integration and validation

[change]

Hi All,
I’m currently working on development of HPLC method for assessing purity of active compound across purification steps (in process control). The method will only quantify active ingredient. The crude extract is a mix of many compounds and their spectra strongly overlap, so I have multiple peaks and wandering baseline. Well...that is most likely not a real baseline but despite all efforts I can’t get better resolution at the moment. It makes integration quite difficult, especially getting a baseline across all peaks. I have a problem with two minor peaks just before the peak of interest. If I establish just a local baseline for those three peaks (just like in Fig 1, overlaid blank in blue) and use drop method then peak of interest will not be baseline resolved, resolution will be around 1.1 and will fail suitability criteria of “not less than 1.5”. However, if I use valley-to-valley method (Fig. 2) then resolution will be around 1.9 and will pass. So I have a few concerns:

1.If I use drop method and stay with resolution around 1.1 would that be enough for method validation and acceptable by regulators as long as precision and accuracy be OK, despite regulator’s “recommendations” for higher resolution values. Or can I just use specification that “peak has to be visually resolved” instead of specifying a value?

2.I know that drop method would generate smaller error, but what if I use valley-to-valley method – would it be acceptable?

Has anyone validated method by using valley to valley without regulators concerns?

3.Is there any official document or guidelines that would specify which integration methods are allowed to use?

Thanks in advance for any replies.
Fig. 1


Fig.2

[Peter Apps]

By far easier is to set the minimum area to miss the two small peaks and see the big one that you are interested in.

Peter

[change]

By far easier is to set the minimum area to miss the two small peaks and see the big one that you are interested in.

Peter

Thanks Peter Apps for your comment, but I haven't confirmed yet if those two peaks are degradation products so I would prefer not miss them.

[Johnny Rod]

"Visually resolved" won't wash really. I would say you'd integrate V-V so you should also look at your resolution that way and put a number to it.

[Peter Apps]

By far easier is to set the minimum area to miss the two small peaks and see the big one that you are interested in.

Peter

Thanks Peter Apps for your comment, but I haven't confirmed yet if those two peaks are degradation products so I would prefer not miss them.

From your first post; "The method will only quantify active ingredient."

So as long as you have an area and a calibration for the peak of interest, what difference does it make what the other two peaks are ?

Peter

[change]

Peter, you are right, but it might be important for future investigations "for information only".

[carls]

The difference in resolution values between the 2 integration methods appears to be due to 2 peaks being identified as one peak with drop line integration (affects RT and width) while only the peak directly ahead of the main peak is integrated with valley to valley.

Depending on the intended use and expected lifetime of the method I'd be concerned with the robustness of this method. Have you checked other batches of sorbent to verify consistent resolution?

What %RSD value do you need for this method? If its the typical 2% you may be headed for trouble.

[change]

The difference in resolution values between the 2 integration methods appears to be due to 2 peaks being identified as one peak with drop line integration (affects RT and width) while only the peak directly ahead of the main peak is integrated with valley to valley.

Depending on the intended use and expected lifetime of the method I'd be concerned with the robustness of this method. Have you checked other batches of sorbent to verify consistent resolution?

What %RSD value do you need for this method? If its the typical 2% you may be headed for trouble.

Carl,
I've check so far just 1 batch so I will have to look at other batches. Thanks

%RSD would be below 4-5% (2% for Reference solution)

[KM-USA]

Isn't the primary cGMP regulatory interest in the finished product? And that any intermediate checks along the way are for internal investigations only?

[change]

Isn't the primary cGMP regulatory interest in the finished product? And that any intermediate checks along the way are for internal investigations only?

Each purification step has to meet purity specification, and this method will evaluate the purity after each step, so this will be crtitical process information.


Has anyone validated HPLC method with valley-to-valley integration?

[Johnny Rod]

Steps where significant fragments of the molecule are added, or those that add or remove impurities, are considered critical process steps in API manufacture from a regulatory point of view.

Re. the size of the peaks, if these end up being greater than the threshold amounts for an impurity profile (e.g. 0.1% depending on the various factors) then you'll need to provide info about them, so being able to get reasonable integration isn't such a waste of time as it might appear.

[danko]

In my humble opinion valley to valley integrations are silly – anyway in cases of baseline isn’t the same as the valley – signal wise. In this particular case would anyone integrate the peaks prior to the one of interest (between 3.2 and 4.5 min) using valley to valley integration method? And if so, what would be the underlying area (the area between the actual baseline determined by injection of a blank and the imaginary baseline created by the valley to valley integration) viewed as? It is the same for the peak of interest and the couple of partly resolved small peaks.

Best Regards

[lmh]

I'm not in a regulated environment, but I really, really think you're worrying about the wrong thing. With that many peaks, that close, how sure are you that there is no chance of anything ever coeluting with your peak?

My feeling is that you have an assay that's useful for information only, but I wouldn't stake my life on the result being super-accurate because there's just too much chance that other chemicals are, to some extent, overlapping with your peak (after all, the fact you have no flat baseline anywhere is telling you that there is always some chemical or another or several eluting throughout the entire region of interest. So a few percent error on how you integrate the peak rather pales into insignificance.

This is the sort of situation that's crying out for more specificity.

[LC_labrat]

2.I know that drop method would generate smaller error, but what if I use valley-to-valley method – would it be acceptable?

Has anyone validated method by using valley to valley without regulators concerns?

3.Is there any official document or guidelines that would specify which integration methods are allowed to use?

USP allows a peak to valley ratio when two peaks are not baseline seperated. USP35NF30 "The p/v may be employed as a system suitability criterion in a test for related substances when baseline separation between two peaks is not achieved. Figure 2 represents a partial separation of two substances, where Hp is the height above the extrapolated baseline of the minor peak and Hv is the height above the extrapolated baseline at the lowest point of the curve separating the minor and major peaks. p/v = Hp/Hv"

As far as I know all conventional integration methods are suitable for use valley to valley, drop line, tangential skim, etc. Other than manual integration which is acceptable only as an occasional last resort and with strong justification.

[prepcolumn]

Change

Imh is right..For regulatory concern; if i were u, i would wonder about proving the purity of your peak in such a crowded matrix sample instead of integration differences..

For integration; i would integrate drop baseline and then split these two peaks and give p/v ratio as system suitability parameter..