system suitability replicates

[rgamage]

Adding some more information to my query.

Method is not validated. Method is not for testing API or product. It is to check leachables from container/closure into product. Work is at product development stage.

USP requires 5 (or 6) injections with %rsd NMT 2%. If my solution is unstable and run time is long, can I cut down system suitability injections to less than 5 and use mean and %rsd to check system and calculate results.

Product is not in USP. I mean USP <621> says SS should have 5-6 injections std mean peak area. How critical is this? I agree MEAN of higher number will be closer to true value, but if I know std solution is not stable, longer I run, it will get fruther from true value. Is it acceptable to do less than 5 injections and take the mean for calculations?

[grzesiek]

what does your client (or you) require?

"If my solution is unstable and run time is long, can I cut down system suitability injections to less than 5 and use mean and %rsd to check system and calculate results." - man how do you live with that? how unstable is your solution? how long is analysis taking?

[tom jupille]

The short answer is "NO".

The somewhat longer answer is that USP methods are, by definition, validated. In order for your results to be valid, the method must be run *exactly* as written. If you change anything outside of the accepted adjustment limits, then you have a different method and must validate that method. (and the number of replicates is *not* an accepted adjustment!!).

As grzesiek implied, the repeatability requirement is intended to flag exactly the type of problem you are seeing!!

[krickos]

Hi

Generally I would agree with Tom any day in the week but I/we have stumbled on a few "exceptions from the rule".

First, USP/EP does not generally state the stability of soulutions(standards/sample) even if you have submitted that validation information, which is "#¤"#¤ so even if validated all important info is not in the monograph

Second, what we have had in the past is sensative samplem solutions that has to be prepared just prior to injection (cooling vials to 4°C in autosampler may help stability) due to quick degradation.

So what I am trying to say is that it may be so that there are a few (old??) monographs out there where the general SSTs may cause troubles. So I would recommend reviewing your experiments to rule out any errros and then contact the contact person (should have acess to validation) for the monograph in question and see if this problem is know for the monograph in question.

In our case the issue was resolved fairly quickly as that monograph was under revision and the new sample preparation gave more stable solutions.

[tom jupille]

Product is not in USP. I mean USP <621> says SS should have 5-6 injections std mean peak area. How critical is this? I agree MEAN of higher number will be closer to true value, but if I know std solution is not stable, longer I run, it will get fruther from true value. Is it acceptable to do less than 5 injections and take the mean for calculations?

There's a little bit of confusion here. Is this an existing, validated method that you are running and having problems with (that's the assumption that I, and I believe, others having been working under) or is this a new method that you will be validating (or is it a "quick and dirty" method that will not require validation?

If it's an existing, validated method, then the solution stability issue should have been addressed at or before the method was validated. Follow krickos's advice and get more information.

If you will be validating the method, the krickos (again) makes a good that there *are* exceptions to the general rules. The best approach would be to deal with the chemistry that's causing the solution stability issue in the first place. If that's not feasible, then putting limits on the allowed time from sample prep to analysis may be required, and you may have to prep 5-6 samples to get your statistics.

If it's a quick-and-dirty method, then do the best you can and don't worry about it.

While the mean and standard deviation of 5-6 replicates may seem like a meaningless exercise, it is very important. To grossly oversimplify: the confidence interval of the analysis needs to be tighter than the control limits on the product (e.g. if your manufacturing specification is that a tablet contains 98 - 102 mcg of API, then you need to have an analysis with a confidence interval of 1 mcg or so).

[grzesiek]

Tom's right

as I asked before : "what does your client (or you) require? " - which means - what is this method for? Who is it for? what do you want from it?