Bioequivalence and Bioavailability Forum

Dear Jaime!

❝ If the drug is subjected to polymorphism, a BE study should be performed

❝ in geno-/phenotyped subjects, if:

❝ - a parallel design is used (e.g., for drugs with a long half-life);

❝ groups should be stratified for the respective geno-/phenotype.

❝ Example: if 20% of the population are slow/poor metabolisers and 80%

❝ fast/extensive ones, both groups (treated with either test or reference)

❝ should consist of the same percentage of SMs/FMs. Otherwise it would be

❝ impossible to calculate the treatment effect properly.

Just a few comments. IMHO it's no good idea to stratify groups for the phenotype, because drop-outs will change the SM (slow metabolizers) / FM (fast metabolizers)-ratio in such a way, that the treatment effect (which is based on group's means) will be biased.
Expanding your example:
Parallel design (2 groups of 30 subjects each; 24 FM (80%) and 6 SM (20%) in each group), responses (FM = 1, SM = 10)
Complete data set: GM_T 1.58, GM_R 1.58, GMR 100%
1 Drop out (Reference group, SM): GM_T 1.58, GM_R 1.49, GMR 1.07%
1 Drop out (Reference group, FM): GM_T 1.58, GM_R 1.61, GMR 0.98%

Therefore IMHO mixed groups should be avoided; only FM should be studied instead.
If stratified groups are to be used (i.e., due to a regulatory requirement), I would suggest laying down a procedure in the SAP excluding a subject of the same phenotype of the repective other group in a predefined blinded manner in order to keep the SM/FM ratio balanced. However, such a method is suboptimal.