Hello all, In the Crossover module, non-parametric analysis can be performed on simple 2-way crossovers. I was wondering if this can be set up for higher order crossover studies. I am interested in comparing median differences in Tmax between 2 different tests and 1 reference (A vs C, B vs C). Thank you! Mike
Hi Mike, No, you cannot modify that module for higher designs. It was based on a paper that only dealt with 2x2. You would need to consult a statistician for an appropriate methodology for your design and median difference but I do not believe you would be able to do it in WinNonlin. Ana
Hi Mike, agree with Ana – no way in PHX. For the theoretical background of a 3×3 cross-over in 6 sequences (Williams’ design) see: Willavize SA and EA Morgenthien Nonparametric confidence intervals for Tmax in sequence-stratified crossover studies Pharmaceut Statist 7/1, 9–19 (2008) Duchateau L, Janssen P, Straetemans R, and C Otoul Adjusting pairwise nonparametric equivalence hypothesis tests and confidence intervals for period effects in 3×3 crossover trials J Biopharm Stat 12/2, 149–160 (2002) You can implement the method in SAS 9+ (Proc npar1way) or one of R’s packages for nonparametrics (wilcox.exact, coin).
Ana/Helmut - thanks for the input and references.. Mike