Bioequivalence and Bioavailability Forum

Hi Achievwin,

❝ Is there any SAS or R-Macro or program for performing RSABR in a 3-period 2-sequnce balanced replicate design?

Use the SAS code for the 4-period 2-sequence replicate design given in the Progesterone and ANDA guidances. Since you have also two sequences, only trivial modifications to select the treatments in the first part of the code. My guess for TRR | RTT (I don’t speak SAS):

data test1;
  set test;
  if (seq=1 and per=1);
  lat1t=lauct;
run;
data test2;
  set test;
  if (seq=2 and per=2) or (seq=2 and per=3);
  lat2t=lauct;
run;
data ref1;
  set ref;
  if (seq=1 and per=2) or (seq=1 and per=3);
  lat1r=lauct;
run;
data ref2;
  set ref;
  if (seq=2 and per=1);
  lat2r=lauct;
run;

Also doable in R (I think Detlew posted sumfink back in the day; too lazy to search).

❝ or how a doubt given below can be clarified?

❝ A 3-period, 2-sequence (TRR or RTT) is proposed assuming it is a balanced design! how to clarify the reference-scaled average bioequivalence approach will be applied to the proposed sequences.

I guess you are referring to the draft guidance of December 2022. You can use also TRT | RTR (see page 7). Accepting 3-period 2-sequence full replicate designs is a step forward because when using the SAS code for the lousy TRR | RTR | RRT (ABE if s_wR < 0.294), the optimizer may fail to converge (see this article).

❝ Thinking Rationale: was that the design was preferred in the current statistical approaches guidance current SAS macros ignore any missing sequence and provide RSABE conclusions with existing macros and secondly any unknown carry-over effect is best addressed in the current sequence by balance effect.

Not sure whether I understand you correctly. What is stated in the guidances about carry-over is nothing more than . Unequal carry-over – which would bias the treatment effect – cannot be ‘corrected’ statistically and only avoided by sufficiently long washouts (see this article).