Chromatography Forum

Is changing the gradient HPLC method (one section in the gradient by =/- 1-2 minutes) a robustness test? I thought robustness tests were only for realistic changes that could occur in your system like temperature, pH, injection volume, etc..

Hi

Generally speaking:

When it comes to the pharma industry ICH Q2 R1 defines Robustness as:
"The robustness of an analytical procedure is a measure of its capacity to remain unaffected by small, but deliberate variations in method parameters and provides an indication of its reliability during normal usage"

Its a quite wide definition, then ICH list some examples like the one you mention but that does not necessary limit the parameters that could need some investigations.

Looking in the pharmacopieas USP (chapter 621), Ph Eur (chapter 2.2 something chromatographic techniques), there are some listings on typical variation allowed in monograph. However neither of those mention gradients as I recall, at least Ph Eur recognises that different instrument set ups can cause problems.

So in conclusion, yes I would say small variations of the gradient falls under robustness, usually you try this during development so perhaps you already have the data?

thanks for your input. we have not done this yet so i do not have any data. someone suggested to change the gradient as a robustness test, but i have not heard of changing the gradient for a robustness test (to me that is changing heart of the method- not a small variation in parameters), but like you said the definition is wide for robustness tests. i guess we will see how it turns out!! thanks again.

Changing the gradient time can be a simulation for the differences in void volume time, so if you change from instrument to instrument, this can be usefull

Ace

We always based it on allowable composition variation, and varied ramp rates accordingly; though I never had to validate a multistep gradient. If you look at the end point of the gradient and variance in time you know what the composition should be and what it would be with simple calculation. The allowable variance should be known from method development, depending on how much time for development you had, but this is NA for retrospective validations (though history may be useful in that regard).

For example, if you have a 30 -> 80% gradient you can run 27 - > 83 and 33 -> 77 and verify performance of the method. I've killed more columns that way.