by
Dan » Thu May 10, 2007 2:50 am
John,
I am not familiar with the TWP3. However, I can comment on the means to compare results for automated versus manual results.
Basically, you are just comparing results of two methods. Typically, the comparison of methods would be treated as you would the reproducibility (precision) validation experiment.
How you perform the reproducibility experiment and apply acceptance criteria will depend upon your relavant SOPs. Generally, the experiment is performed by generating six replicate sample results for each method. The results are evaluated and statistical data is calculated. You can then apply limits to: the difference in means (relative or absolute difference), individual RSDs of each data set and overall RSD of both data sets. There are more advanced statistical data that can be generated. I prefer the simple ones that I noted here. I have seen various statistical test such as the f- and t-test used, but genereally, those type of tests are not appropriate for the small data sets that are generated in a method validation experiment.
Also, there may be portions of your automated method that need to be evaluated in the validation. These are generally procedures or that are not present in the manual method and, thus, were not validated previously. These things may not need validation but they may need optimization and that can be done as part of the method validtion. In your example, you will be transferring samples from the TWP3 to and HPLC. Questions to ask: does the transfer need to be validated?, is there a filter used in the transfer that needs validation?, etc.
Regards,
Dan
