Bioequivalence and Bioavailability Forum

Dear Ben,

I must confess that I haven't dealt with sample size planning to a given precision of the confidence interval up to now. Within that forum this type of sample size planning is a little bit Off Topic.

All sample size planning within BE studies is usually done via power of the underlying two one-sided t-tests (TOST) for the bioequivalence decision. That's the reason I had written the R package PowerTOST.
Cool that you have discovered some hidden helper functions and tried to reuse them . But beware! These functions are intended for internal use only and are therefore undocumented. Using them needs an understanding what they do and of the assumptions underlying the returned results.

So in the end I can't say something specific to your problems without delving deeper into the mathematical apparatus of sample size planning with the aim of attaining a given precision. For that I haven't enough spare time in the moment.

Only some remarks and tips:

• I have found it always better to be doubtful first to the own implementation than to question the results of others.
  
• If you don't trust commercial software (of course there may be bugs or some erroneous implementations as we all know) try to get some results or test cases from other sources. For your problem I recommend you some sources written by the "pope of sample sizes"
  
  S.A. Julious
  "Sample sizes for clinical trials"
  Chapman & Hall/CRC
  Taylor & Francis Group
  Boca Raton 2010
  
  S.A. Julious
  "Tutorial in biostatistics: Sample sizes for clinical trials with normal data"
  Stat. Med. 2004, 23, p1921-1986
  
  Both contain worked examples and tables of sample sizes you can use to check your implementation.
  
• Have a look into Dave Dubbins FARTSSIE.xls (it's free) which contains sheets dealing with your problem.
  
• The degrees of freedom and the so-called design constant bk you are reusing are defined within PowerTOST for the parallel group design based on the number of subjects in one group assuming equal number in both (i.e. n_total=2*n).
  For the crossover designs they are defined in terms of the total number of subjects assuming equal number in each sequence group.
  So check what your implementation will give or needs. Check also what nQuery gives.
  
• I'm not quite sure if the design constants are specific for the TOST procedure and make sense for your problem. Check the underlying formulas for each design you will implement with respect to these constants.

Hope this helps.