## FDA RSABE is ISC [R for BE/BA]

Hi Helmut, Hi ElMaestro,

» ...
» Well, the EMA’s models are wacky because they assume equal intra-subject variances of T and R. An assumption which was shown to be false in many cases. Essentially most of the information obtainable in a replicate design is thrown away.

Full ACK!

» I don’t know why the FDA does not directly estimate $$\small{s_\textrm{wR}^2}$$ from the mixed model but work with the intra-subject contrasts of R instead. ...

I was always wondering about the different methodologies in the code of the progesteron guidance:
• ISC analysis for estimating CVwR and decision for using RSABE or ABE based on this estimate
• Mixed model analysis of the 90% CI of T-R (ABE analysis) in case of swR < 0.029356 (CVwR < 30%)
• ISC analysis for estimating the point estimate of T-R including 90% CI and its use in the RSABE criterion in case of swR >= 0.029356 (CVwR >= 30%)
Why not use the ISC estimate of T-R in the ABE decision in case of swR < 0.029356 (CVwR < 30%)?
Politics? Nostalgia (Since years recommended the Proc MIXED code)?

Or the other way round: Why not use the components of the RSABE criterion (pe and 90% CI, s2wR) from the mixed model approach?

I would opt for the full ISC approach because it may be unambiguously implemented in SAS, R, Phoenix and so on for every replicate design, full or partial .

Regards,

Detlew