Bioequivalence and Bioavailability Forum

Dear All!

News from Rrrr land. In the recent past some interesting R packages were released which allows to do the BE analysis more strictly along the way SAS initiates do:

1. Package lsmeans, author Russell V. Lenth (Yes this guy who abandons post hoc power), gives the R adepts access to the LSMeans with ease so that µT-µR and its confidence interval can be obtained via difference of LSMeans. Not really necessary but cool .
   
   
2. Package lmerTest brings the Satterthwaite's and Kenward-Roger approximations for testing fixed effects into the analysis via mixed models. The Kenward-Roger method in lmerTest is taken from package pbkrtest.
   Have also a look at package varComp.

So far so good. Unfortunately these last packages are written for mixed model fitting via function lmer() from package lme4. This package in contrast to the older nlme does not allow to specify different residual errors for groups. Quotation from the lme4 manual: "... lme4 does not currently implement nlme’s features for modeling heteroscedasticity and correlation of residuals...".

The effect of this restriction is that we can not mimic the SAS Proc Mixed code given in the FDA guidance "Statistical approaches ..." for analyzing ABE in replicate crossover designs .
But eventually there is hope considering the word "currently".