System Suitability per USP <621>

[LShelton]

Involved in a bit of a debate over the system suitability requirements of USP <621> with regard to continuing calibration standards in an HPLC run. I interpret USP <621> as requiring EXACTLY five repicates for the calculation of RSD if the specification is </= 2%. The SOP for the lab states that AT LEAST five replicates are to be used. In our SOP, five replicates are used for the initial determination of system suitability, but to determine that system suitability is maintained throughout the run, the entire sample set of continuing calibration standards is used to determine the RSD. This can be as many as 20 replicates.

My argument is that an outlier can be masked by using too many replicates for this calculation and only five standards at a time should be used to calculate RSD at a particular point in the run. Alternately, I'd like to see deviation from the mean used to determine if a particular calibration standard is out-of-trend.

Thoughts??

--Lorraine

[tom_mizukami]

There is some additional guidance in this document: http://www.fda.gov/cder/guidance/cmc3.pdf

In that document they recommend RSD</=1% for n>/=5.

I think the intent is clear, demonstrating that the system was suitable at the time of analysis. I couldn't find the exact reference and it may have been superceeded, but some guidance also mentioned three up front injections and two at the end.

My favorite scheme is just five interspersed standard injections. Calibrate and system suitability precision off of the same injections. Either the whole run is accepted or the whole run is discarded. What happens if you pass up front suitability but fail ongoing suitability? Then you get into a mess of what injection to keep and what injections to discard. I would rather have validated 72 hr sample stability and simply fix the instrument and reinject.

[Tim]

We routinely do 5 injections for RSD as part of the initial system suitability tests for each HPLC run (i.e. new mobile phase, column, HPLC parameters). This has to be <=2% to pass.

During the run we intersperse standards (every 2 for duplicate assay, every 5 for CU and every 6 for Dissolution). At the end of the run we do an evaluation where all the main peak areas in the standards have to be within +/- 2% of the mean of all the std areas. Depending on the circumstances, we either reject the whole run if any fall outside this, or accept the run up to the last good standard before the one that was out.

[LShelton]

Thanks, that's what I was looking for: what are other folks doing to evaluate continuing calibration standards? I like the method described of +/- the mean of the initial five replicates to flag outliers. Anyone using anything else?

[Basil]

Hi

In the same way as Tom:

We routinely do 2 blanks (solvent or mobil phase), 5 injections for RSD as part of the initial system suitability tests for each HPLC run. This has to pass the requirements (resolution, RSD, Peak tailing, ....)

If there is more than one batch or the system goes longer than 12 hours, we add at the end of the sequence 2 more injections and compare with the first 5 injections. If the last two replicates are wrong i don't belive all the analysis and we did it again....

When we pass the FDA inspection they comment that to control system suitability injections should be repeated consecutivly...

I hope this help