"k' > 2" is a
suggestion in the system suitability section of the FDA "Reviewer Guidance" from 1994 ( ! ) :
https://www.fda.gov/downloads/Drugs/Gui ... 134409.pdf . It is only a suggestion, so that a validated method with k' < 2 *should* be OK -- assuming you can get a reviewer to overcome his/her bias (see this post for a take on the subject:
viewtopic.php?f=1&t=48623#p233084 ).
That said, k' > 2 is a very good suggestion
most of the time:
- a low k' hurts resolution from nearby peaks [Rs is roughly proportional to k'/(1+k') ]
- anything that is unretained elutes at k' = 0, so low-k' methods risk being non-specific
- the equilibrium of the chromatographic system is disturbed by injecting a sample and it can take a finite time for the system to decay back to equilibrium; trying to quantitate on a dis-equilibrated system is error-prone
There are a number of "gotchas" in those arguments. Most important is the fact that while k' is well defined k' = (tr/t0)-1, the dead time (t0 = elution time of an unretained compound) is
not! In the extreme case of size exclusion (GPC or GFC),
all of the peaks have k' < 0 ! Even measuring t0 is can of worms, as attested to by this old thread from the Forum archives:
http://www.lcresources.com/discus/messa ... 20041222pm