Bioequivalence and Bioavailability Forum

Dear all,

following this example some more remarks. Sun et al.¹ proposed three ‘classes’² of interactions:

• concordant quantitative: the treatment effect is overall equivalent as well as in all groups but differs in magnitude,
  
• concordant qualitative: the treatment effect is overall and in at least one group equivalent, in at least one group not equivalent, and the treatment effects in all groups are in the same direction, and
  
• discordant qualitative: the overall treatment effect is equivalent, the treatment effect in some groups is not equivalent, and the treatment effect in some groups can be in opposite directions.

The wording ‘equivalent’ is somewhat misleading, since the method assesses only the point estimates. Therefore, it is possible that – although both groups fail BE with the confidence interval inclusion approach – the interaction is still classified concordant quantitative. That makes sense because the power of model III of groups is low and thus, assessing the confidence intervals is too restrictive. If an assessor expects that all PK metrics of all groups pass conventional BE, why would the applicant have been so stupid as to conduct such a large study?

With the example of the linked post:

• C_max: overall by model II 107.19% (90% CI: 100.55–114.26%) ✅
  Group 1: 117.10%, Group 2: 99.14% → concordant quantitative interaction ✅
  
• AUC: overall by model II 101.15% (90% CI: 94.34–108.45%) ✅
  Group 1: 108.01%, Group 2: 95.34% → concordant quantitative interaction ✅

See also the results of our meta-study, where all interactions were also classified concordant quantitative.

---

1. Sun W, Schuirmann D, Grosser S. Qualitative versus Quantitative Treatment-by-Subgroup Interaction in Equi­va­lence Studies with Mul­tiple Subgroups. Stat Biopharm Res. 2022; 15(4): 737–47. doi:10.1080/19466315.2022.2123385.
   
2. My terminology.