Purity factor threshold

[Dominique]

Hi everyone!
I am using an Agilent 1100 with DAD and need to do some peak purity analysis. The issue is when to consider a peak pure based on the purity factor, since i have observed that the calculated threshold limit varies a lot between injections and also between analytes (from 996 to 999).
What criteria should I use to determine an appropiate threshold if I decided to set it manually?
I'd appreciate any advice on this subject...
Thanks!
Dominique

[pliva]

I nornally put 999 and its accepted.

[HPLCCONSULT]

Please do not take "pliva" advice without understanding the feature first. There is no universal value and to just place a value in the box without considering what you are doing is just bad science.


"Peak Purity" by UV/VIS is a very complex issue. You must take the time to educate yourself (or better yet, through a proper training course on HPLC Diode Array Optimization) about all of the various options, functions, parameters and settings in the Peak Purity Menu Screens. Additionally, you must have a robust method to start with where the data have been correctly gathered at the optimum wavelength(s), data capture rate and noise levels with a general understanding of what the structural similarities of any possible "impurities" might be. Impurities with similar structures may appear to not be impurities at all. *Number one problem we see is when someone obtains "purities" where the HPLC method used is really of poor quality to begin with. Time must be spent developing a proper scientific separation method with optimized parameters BEFORE using the feature.

The Software menu's in ChemStation (And this is true for most brands) allow you to define the "Purity" threshold at any value you want. '999' is not a good choice to use without understanding what you are doing and what the limits of your method are. The scale is from 1 to 1000 (to replicate a scale of 1-100% in tenths). The ideal value to use would be one that has been obtained through actual analysis of pure samples using HPLC and other complementary techniques (e.g. MS, NMR). These would then be part of an SOP defining ALL of the HPLC parameters for the analysis plus ALL of the 'Spectral Purity' software values (and hardware values such as flow cell path length and volume; slit width, bandwidth and so on). *** Keep in mind that samples can appear to be "pure" when the impurities present have similar absorbance properties.

"Peak Purity" can be a useful feature, but in my mind it is a very advanced function which is often abused by chromatographers. It should only be used by someone with an in depth understanding of the HPLC parameters and acceptance of the qualitative nature of the result obtained.

[mbicking]

I would like to start by agreeing, and also diagreeing, with the two previous posts.

It is true that the Agilent Peak Purity software is complex (I have a 40 page report on the statistics involved, and it is at my limit of understanding). However, if you really understand the Peak Purity software, you will probably decide not to use it! The problem? I can make the peak pure or impure (green or red on the screen) just by adjusting the parameters.

This approach has been developed to produce a simple answer ("Green is good, red is bad") to a complex question (Is this a pure peak?). It allows less experienced analysts and decision makers to draw conclusions without thinking about what the data really say (the computer says it's pure so it must be true!). A false sense of security, I think.

If you have poor sensitivity and poor chromatography, you are likely to get an unreliable result, which makes sense. But if you have very good chromatography and spectra, the system will often label the peak as impure because the noise-based threshold is 999.7 and the actual peak is at 999.6. This, of course, is an absurd conclusion.

Using a value of 990 as a fixed threshold is a good place to start. But I also agree that you must not stop there. You must look carefully at the data, and all the other information that you may have, and then draw your conclusions based on the science, not the color on the screen.

[HW Mueller]

We have been through this many times. From these discussions and my own experience with UV spectrometers (outside of chromatography) and multiple wavelength detectors one can condense the following: If the spectrum does not look like it should one can conclude something is wrong. If the spectrum is as expected for a pure compound then you know practically nothing, since, as has been mentioned here also, you don´t know whether you have different compounds of identical UV specs. I don´t expect this to be very rare, because of a tendency that compounds with the same spectrum are structurally close and elute together.
This is why our teachers never accepted chromatography, LC-UV or UV for proof of a compound.

[HPLCCONSULT]

This is why science teaches us to use multiple techniques to determine "purity".