## CV values in BE studies: intra, or inter, or total? [Study As­sess­ment]

Hi Elena

perhaps not a direct answer to your actual question but a general comment to the general question you aired:
1. For a crossover trial with 2 treatments, 2 sequences and 2 periods we use the "within subject variance" to construct the confidence interval. The CV is derived from this variance. We can call it a within-subject CV but it is not treatment-specific; you can say it is an estimate of a common intra-subject variance or CV. CV = sqrt(exp(Variance)-1). We get that variance as the residual error from a linear model with various factors.
2. For a crossover trial with replication of the reference treatment the important CV, but not the only one, is the CV associated with the reference treatment.
3. For a parallel trial we derive a total variance and total CV for the construction of a confidence interval. If the trial is done in batches, then the CV is somehow a kind of total CV excluding the component of the influence of batch, if calculated correctly. Yes, cosmic mindf%cker but BE often is.
When you read study reports or assessment reports you often come across descriptive statistics. Here the CV is simply the (sample) sd divided by (sample) mean.

If you want to use a CV for sample size calculation, then you need to know which model you applied in the dataset where the CV came from and what the model is the design you want to apply - the model is dependent on the study design itself. If you have a CV from a 222BE trial, then obviously this could be directly used for sample size calculation in another 222BE trial assuming your assay and other factors are as good as in the other trial. If you want to apply the CV from a 222BE trial to a replicated trial then that is still a decent guess in my opinion and vice versa. What you should not do is to apply the CV from descriptive stats to any sample size calculation, or to confuse CV's from parallel trial with CV's from crossovers.

Intra-subject CVs are generally (=most often) lower than between-subject CVs which are lower than total CVs.

Pass or fail!
ElMaestro  Ing. Helmut Schütz 