Number of replicate injections for a sample

[Kimico]

Hello everyone,

All the LC separation methods that we currently use call for 2 or 3 replicate injections of the test sample (3 replicates/same vial). It seems to me this is a waste of resources due to the superior performance of today's autosamplers and the fact that we run system suitability at the beginning of the run and every other 6 samples.

I brought this to the attention of our director and my plan was to do control charting to demonstrate that since the implementation of the method, the autosampler's precision has always been below 0.5%RSD (Specification for the method is NMT 2%) for the given compounds we analyze with that method, and only in very strange occasions you get a weird injection. All this with the intention to reduce the number of injections from 2 to a single injection.

My questions to everybody are the following:

1)How many replicates for a sample (per vial) do you usually do in your labs?
2)If you run one single injection, How do you determine when a given injection is outside the performance limits?

FYI, I work in a regulated environment (FDA) and thanks anyone for your input.

[lmh]

I work in an unregulated, academic environment (much easier!)

We differentiate between analytical and biological replicates. It is absolutely essential to have proper biological replicates (i.e. if you study diseases of raspberry, you need to prepare samples from many individual diseased plants, and many individual healthy ones). Given that biological variation is usually vastly greater than analytical variation, it is generally unhelpful to include analytical replicates.

Having said that, it is useful to include QC samples (and I generally use multiple calibration curves) which would reveal a gross failure of autosampler.

A VERY big gripe of mine is finding groups that do not differentiate between types of replicate, and run three biological replicates three times, and regard this as n=9. It is not. But you can make the SE look much smaller that way...

The important difference between the academic and the regulated environment is what we're measuring. Academics usually want to know the mean value of something ("does X increase when raspberry plants gets diseased?"). The regulated environment usually wants a reliable value for a single sample (is this batch of drug safe to use? Does that patient have kidney failure? Has that athlete got a significant chemically-induced advantage in the 100m sprint?). Hence academics don't need to worry as much about the validity of individual measurements provided the overall mean is correct.

[LC_labrat]

I also work in a regulated environment (FDA/GMP). I've typically ran test samples with duplicate injections and a 2.0% RSD suitability requirement for the sample. It's much easier to prove a bad injection by not meeting the requirement and re-inject than to have go through a potential out of specification investigation.

[HW Mueller]

Since one of the reasons for existence of academia is to gather knowledge it seems that one has to impose very stringent checks/controls on oneself, in any case much more effective controls than regulation requires.

[krickos]

Hi

First off, yes I am in a regúlated area too (EMEA/FDA....) with regards to APIs.

We abonded the duplicate injections for a vial some time ago for LC but for example we still keep it for CE injections though.

Breifly I recall that we did some risk assesment from several angles but in short:
-SSTs. The general 2,0% limit for RSD is nonsens (ie something is very wrong if RSD exceeds 1,0%) with regard to our applications and instruments performance. Consequently we apply a 1,0% limit instead. And we also tend to "bracket" stds/SST vials for RSD purposes.

-With each batch there are at least 2 samples=> a minium of 2 sample preparations/vials. If only one sample (stability sample for instance)=> two preparations/vials from that sample. Adresses question 2 i some ways ie the API batch should be homogenic (ie the results are compared) however not sure this approach would be suiteble for all drug products depending on formulation (tablets vs bottled liquids?).

- Of course we have reviewed the history of instrument problems (and countiously do), and injection related issues are very rare compared to, poorly washed/conditioned system, column degradation, user mistakes ecetera causing primarily other SSTs than RSD to fail.

So in conclusion, I suggest that you do a risk assesment (you seems to already have started).
Do not only use SSTs as an argument (see above) as it may not reflect you instruments performace levels well enough, in fact FDA/CDER already in 1994 said that less than or equal to 1% limit is "desireble" (page 22 Reviewer Guidance "Validation of chromatographic methods" Nov 1994).