Minimum Theoretical plate number

[Haythem KHALIFA]

Dear Sir or Madam,

I'am asking you hoaw can i fix plate count during validation studies.

Best regards

[dap]

Hello.
You should modify system conditions to increase elution time and/or decrease peak wights.

[tom jupille]

There is no single, simple answer to that question. We would need more detail about what you are trying to do. For example:
- have you already developed the method and are starting to validate?
- or are you developing the method ab initio?
- how many analytes are you looking at?

[Haythem KHALIFA]

It is a well validated method with a single analyte ( Macrogol 4000). The plate has not exceed 1200 even with a new column. It 's GPC methd wi RI detector.
How can' i set a minimum plate count number in the system suitability ?
The 1994 FDA guide FDA « Reviewer Guidance Validation Chromatographic Methods» said " The theoretical plate number depends on elution time but in general.should be > 2000 ”. I think that we can do exception.
In addition supposing that we have a value of N=4000 during validation, i suppose that i should fix a minimum of 2000 in the system suitability. Is it true or false.

[Perreman]

Well since you are using GPC it is a different story. I don't think the recommendation of >2000 plates includes GPC analysis.

Regarding your question of setting a limit: The proposed plate limit in your method should be based on previous knowledge (ex. validation study, robustness testing). However, an ideal limit should be set by knowing when the chromatography is disturbed to such a degree that a result is valid or not.

To clarify: You should study the methods performance for each SST parameter and set a limit based on this data.

[dap]

Symmetry & Plates shouldn't be applied to GPC because of analytes nature.

[mattmullaney]

Good Evening,

I'll note another excellent example of a stationary phase that does not afford high plate counts...the AGP (protein-based) chiral phase made by Daicell and others...this phase is historically poor and noted in pgs. 691 - 696 of Introduction to Modern Liquid Chromatography, 3rd Edition, by Lloyd R. Snyder, Joseph J. Kirkland and John W. Dolan, ISBN: 978-0-470-16754-0. This one came up at work recently...compelled a revision to an SOP I think.

[mattmullaney]

To complete my thought(s), why not evaluate statistically the plate count for the application after say, 30 - 60 injections of real samples and standards, and assign the plate count in that manner rather than "forcing" a number? Or, use other criteria for system suitability?

[tom jupille]

To expand a bit on the previous ressponses:
- the concept of "plates" implies that the dispersion in the peak is due entirely to random variations in migration velocities among *identical* molecules. In GPC, however, a significant part of the dispersion is the result of molecular weight distribution. That's why plate number is not really relevant to GPC.
- that said, plate number is a convenient metric to use for peak width (if for no other reason than the regulatory people are used to it!). As your column ages and the peak shape deteriorates, there *will* come a point where the excessive peak width begins to impact your quantitation. That peak width is the maximum you can tolerate (i.e., that plate number is the minimum you can tolerate). As suggested previously, what you have to do is to go back over your experimental data to determine where that point is (and then build in a safety factor!).

[Haythem KHALIFA]

Thanks for the responses.
On the other hand i noticed that the european pharmacopea and the USP disregard in many cases the plate number. In some monograph they impose only area RSD. Is that a result of robusteness study ?
Can ' i use the column until N value equal to 500 ?

[dap]

N and A mostly depends on polymer fractional composition not on column health. To check column health polymer calibration mixtures should be used. We use dextrans f.e.: https://crs.edqm.eu/db/4DCGI/search?vSelectName=1&vContains=1&vtUserName=dextran

[mattmullaney]

Hi Haythem,

It's hard to say from where the pharmacopeias' decision is being made not to use plate count as a system suitability criterion--that said, what Tom Jupille and Dap say makes a ton of sense to me...the nature of the sample(s) and the means of separation simply do not lend themselves to using plate count in this manner.

It's kind of hard to know how long you can use the column...without a series of chromatograms and a statistical history of their plate count(s).

Are peaks nearly co-eluting? What is/are the resolution values of the peak of interest from its nearest neighbors? At what point is the peak of interest becoming difficult to integrate?

I noted before...it is not likely the FDA considered all types of HPLC separations in making the "blanket statement" regarding plate count back in 1994 or so. My current firm also made the error of going along with this pseudo recommendation (I recall the word "should" was in the FDA document w.r.t. plate counts "should be >/= 2000:=")...yes, we were bitten by this as well as we failed to think before we leaped in following this idea.

If you have, say, 30 - 60 injections of this analyte and the corresponding plate counts, why not draw the arithmetic mean and standard deviation, multiply the standard deviation by three, subtract that figure from the mean, and use this as a low-end for plate count?