Chromatography Forum

I'm a consultant doing work at a medical device company, and I need to find a resource that explicitly states when and how system suitability testing is to be performed. Does anyone have ISO 9000 or 13485, or FDA documentation on establishing and verifying system suitability for laboratory test methods?

This may help a bit.

http://www.fda.gov/downloads/Drugs/.../ ... 134409.pdf

I do not think you will find a document stating "you must perform x and y" for every type of analysis. Although I do feel there are some core system suit criteria such as %RSD and Tailing Factor. Usually system suitability is derived from experimenting and finding key points that are crucial to the individual method you are using such as resolution between two peaks.

I was hoping for something more concrete, but I think I found the best guidelines I could in USP 621. I was trying to prove the necessity of concurrent suitability analysis from a regulatory perspective. Nevermind the scientific perspective is glaringly obvious, but that wasn't good enough for the lab manager to want to change the current "once a week" (I kid you not) System Suitability requirement for GCs in the lab. I haven't even tried to look at the LC stuff, and I'm glad it out of the scope of the remediation project I'm working on.

"once a week" doesn't sound that bad to me. But I'm not in a regulated environment.

Depending on the situation, once a week may be fine...
In the end, it's what your SOPs call for and whether you can justify same.

For a dedicated machine testing pretty stable analyte solutions that has a good track record, once a week is probably OK (right up until it is not - you always have to be prepared to remediate bad data w/ fresh standards, control samples, a different system... something, again - better off if covered by SOPs).

I guess if your method(s) and system(s) are really that rugged... however, upon a system suit failure I'd feel that all of your data collected over the entire week since the last system suit was established may be questionable. How many samples would that put into question? Where I work that would be devastating!

I guess if your method(s) and system(s) are really that rugged... however, upon a system suit failure I'd feel that all of your data collected over the entire week since the last system suit was established may be questionable. How many samples would that put into question? Where I work that would be devastating!

True!
But if you're using it for inprocess testing, say for a continuous process, you could establish sys. suitability and just go sample, std every time a sample hits the lab. So long as the std is within spec (dev. from curve NMT 2% or something), you're good and never risk >1 sample. If you're getting loads of samples at predictable intervals, this makes a lot less sense.