Power function: practical and theoretical limits for adding additional replicates to a study planGot a question from a customer in biopharma:
First question is the role of statistical replicates – when using the power function in JMP to calculate “n” based on assay precision and desired CI window, what are the practical and theoretical limits for adding additional replicates to a study plan in order to compensate for the broad windows usually set for bioassays (i.e., 80-120% relative potency usually being accepted as comparable, but adding extra replicates to try to detect a 10% shift in RP)?
I wonder if I have to consider any subject matter expertise to answer this question? Any guidance appreciated.
Thanks,
Martin
Re: Power function: practical and theoretical limits for adding additional replicates to a study planYes, Mark is correct. Sample size and power calculations for equivalence tests will be in Early Adopter release 2, due out early May.
Re: Power function: practical and theoretical limits for adding additional replicates to a study planWell, it is still a simple power calculation but for testing the equivalence of two means. Caleb King in JMP development is creating a new sample size explorer and has included this case as one of the calculations.
Re: Power function: practical and theoretical limits for adding additional replicates to a study plan@markbailey For the last question this might shed some light on it. I will check what they want with their replicates.
The second question is how to use JMP calculate the appropriate number of samples (and their replicates) to compare using 2 different assays in a basic equivalency study to replace one bioassay method with another (such as using ELISA in the place of a cell-based assay).
Re: Power function: practical and theoretical limits for adding additional replicates to a study planthanks, glad my gut feeling that this is probably nothing with a simple statistical answer alone hasn't proved me wrong:)
Re: Power function: practical and theoretical limits for adding additional replicates to a study planI read the original post several times and it is not clear what they are trying to do. I mean, they state that they want to use a power calculation to determine the sufficient sample size, but they don't say for what purpose. I think that the USP and ICH might be better references for standards. CLSI addresses evaluating assays. USP and ICH address evaluating substance and product.
If they are going to collect all the replicates at the same time, by one technician, using one lot of reagents, one incubator... (you get the idea), then they can use the One Sample Mean calculator. If it is more complicated than that, they have to derive an error budget with all sources of variation that will not be constant and compute the standard deviation of one observation.
Also, are they trying to demonstrate a difference or equivalence?
Re: Power function: practical and theoretical limits for adding additional replicates to a study plan@Byron_JMP may have some thoughts...but my initial reaction is "it depends" and then my second reaction is to find a CLSI guideline (probably EP21) that is specific to what they are trying to do and follow that.
Re: Power function: practical and theoretical limits for adding additional replicates to a study planUSP<1033> has a section on assay format for reducing variation in bioassays. I've found that to be a useful reference in the past.
The general idea is to find a combination of run and replicate numbers that are appropriate for the level of variation in the method, and is practical for execution of the method. At the end of the day, its taking advantage of the central limits theorem and using it to your advantage.
