By emmanuel on Monday, September 1, 2003 - 04:56 am:
I would like to know what is the best practice to calculate assay results : is it to take the mean of all the standards thorought the run provided the rsd is less than 2 % or is it to use the mean of the standards that bracket the samples only?
By Anonymous on Monday, September 1, 2003 - 06:26 am:
we use the mean of the standards that brackets the samples. just in case there is case some small changes during the run.
By emmanuel on Monday, September 1, 2003 - 08:48 am:
Thanks for that explanation.
But does this mean that whatever % rsd of the standards thorough the run you get you can calculate the sample results by using the bracketed standards? Or is it part of the system suitability criteria to get less than 2 % for the rsd as a rule of thumb?
I am still confused.
Thanks.
By Anonymous on Monday, September 1, 2003 - 10:34 pm:
In a singele point calibration I prepare duplicate standards and start by injecting 6 x S1. This is initial injection precision and I have a criteria rsd max 2%. Then I run 2 x S2 for standard concordance. After each 4 or 6 sample injection I run a bracketing standard. If I for example have decreasing standard response or rising backpressure I can also set an acceptance criteria for all the bracketing standards throughout the run. Depending on method this can be for example rsd 2%. You can evaluate the value by checking when your sample results gets too much effected. This allows you to reject brackets from the end of the run if you get outside the criteria.
By Tim on Tuesday, September 2, 2003 - 06:54 am:
In our labs we have used a running average through the run, though are changing to using the mean of the whole run.
If any standard peak area falls outside +/- 2% (or +/-5% for low level standards, e.g. degradates) of the mean for all the standards, we (usually) reject the run.
This method shows up outliers better than just relying on the RSD and is quicker than performing true statistical outlier tests.
We have been using this method for as many years as I have been here (<11 years) and have been through many FDA/MCA, etc. inspections with no issues being raised.
By Anonymous on Tuesday, September 2, 2003 - 08:35 am:
I would like to caution you against using a parameter like RSD for all standard throughout the run because it can hide intermittent problems. Take as an example 11 standards total. 10 of the standards have a value of 99.9 and 1 has a value of 95. RSD for the run would be 1.5% but if the assay has a 97-103% specification all points around the 95 are suspect.