I have a QC question about the order of sampling when you are starting a run. I usually run 3 standards to calibrate my instrument then a blank and a hi and low control, then I run 10 samples and then my two controls again. What is recommended for good QC practice when you are running a lot of samples? Also what is a recommended range (ex:+/- 5%) that the control must be between to pass QC? Or is this a part of GLP requirements?

To Jocelyn:

The accepted technique for calibration when running many samples is a technique known as "bracketing". What you need to do is to first run your blanks, followed by your first standard, then your unknown sample, then your next std, and so on. Start with the lowest standard concentration and move up as you go along. For eg. 10ppm, UNKNOWN1, 20ppm, UNKNWN2, etc.

As far as GLP is concerned, you need to run at least five replicates for each sample - standard or unknown. Most pharmacopoeiae specify a reproducibility within +/- 2% CV.

As far as I know, the recommended concn. range for your controls is +/-20% of the label amount. ( I do hope you have an autosampler !!)

Hope this solves your problems. do let me know how you get along.

Warm regards,
srinivas

To Jocelyn,
As far as my experience has taught me for a large run of samples it can depend upon the length of run time.
A std A should be run first then 2 blanks then at least 4 more std A and 1 std B which sould be at leat 25% more in conc. than std A.Then the samples are run. If the run time is quite long then a std A should be run every 6 hours. If the run time is quite short then a std A should be run every 8-10 samples.
I agree that the reproducibility of the std A should be +/- 2%CV.
The std B conc. should also be worked out with respect to std A.
I gained this information from working in GMP and GLP environments so I hope they are of some use to you.

Jocelyn
As far as I am aware there are several ways you can go about this.
As far as System Suitability is concerned.
Five injections of a standard giving an RSD on peak area of less than 1% should be sufficient.
Also a second standard prep in order to confirm your first standard was correctly prepared.
As far as throughout the run:
If you do not know the expected range of your final results, then a pre-run linearity check performed over the entire range of your sample results. i.e 5-150% of your standard concentration.
Or alternatively a three point linearity between every six samples.
If however you are expecting a tight range of results and linearity has already been demonstrated, then just an injection of Std 1 and Std 2 at around the concentration expected can be performed...bracketting as already mentioned.
With the six samples results calculated on both bracketted standards sets.
Dave

I met suitability problem recently. I run two sets of standard 1,2 and 3 before and after samples. Each standard contained 3 components. Everything was fine except component A in standard 1. The peak area counts in std.1 before was only half of the one in std.1 after. That resulted in the correlation coefficient for conponent A failed the criterion of 0.97. This happened to different analysts several times already. However it doesn't happen everytime. I did experiment with a blank after sample and befroe the std.1 after the sample to see if there was a carry-over. I didn't see any carry-over however. This problem bothers me a lot and I can't figure out the reason. Could anyone give me any suggestions? Thanks.

Xiao

What's the retention (k'-value) for component A?

Using Zorbax C18 column, component A almost was the last one eluting out the column with rention time around 29.8min. K' was ~8.7. The retention time and peak shape for component A were consistent throught the run.

To Tom:
I just posted this message by using Creat a New Conversation. This was my first time to use this system and did something incorrectly yesterday. Hopefully see you there.

OK, that eliminates one possibility (if the problem component were eluting too early).

Next thing to check would be to run a series of injections of your standard 1 (say, 5 or 10 in a row) and see if the area for component A increases gradually during this series. If it does, this would suggest sorption on active sites on the stationary phase.

-- tom jupille / LC Resources

The columns used with this queation were not new ones. They have been used for around 5-7 times already, which means they should be equilibrated well if there were active sites on the stationary phase. However, I will try your suggestion to see waht I will get.