Bioequivalence and Bioavailability Forum

Dear Mahmoud and all,

❝ Statistical methodology for highly variable compounds: A novel design approach for the Ofatumumab Phase 2 bioequivalence study

Jones B, Li B, Bagger M, Goodyear A, Ludwig I. Statistical methodology for highly variable compounds: A novel design approach for the Ofatumumab Phase 2 bioequivalence study. Pharm Stat. 2022; Early View 23 May 2022. doi:10.1002/pst.2233.

❝ Used scale average bioequivalence for parallel design

❝ Is this method accepted by FDA and EMA

Huge study: Multiple dose, 284 patients, 30 centers in 9 countries, two-stage design. In the paper we find:

[…] there is no corresponding FDA guidance on BE study designs in a parallel-groups setting for such complex compounds. To overcome this, a mixed-scaling testing strategy for a parallel-groups design in RMS patients was devised. After completion of the study, the FDA, EMA and other global regulatory bodies reviewed the data and accepted the conclusion of bioequivalence based on the results obtained from the APLIOS study, using the proposed methodology.
(my emphases)

Incidentally I know one of the reviewers of the manuscript and he was not the only skeptic. 
In general reference-scaling may be applied for drugs / drug products with a high within-subject variability of the reference (for the FDA \(\small{s_{\text{wR}}\ge0.294}\) and for ones applying ABEL \(\small{CV_{\text{wR}}>30\%}\)). If we would accept reference-scaling based on the total variability, it would open the door for drugs with polymorphic metabolism, where \(\small{\sigma_{\text{inter}}^2\gg \sigma_{\text{intra}}^2}\). Furthermore, it was an adaptive design (stopping in the first stage for success). Since the upper 95% bound of the linarized criterion was used, obviously without α-adjustment. Fine for the FDA but for the EMA? I’ve heard that there was some to and fro in the review process. The authors claimed to have discussed the approach with the FDA beforehand, and it was accepted. I guess that was a rare exception.
I’m somewhat surprised that the study was accepted by other authorities as well (ABEL ≠ RSABE). Although the point estimates of both AUC_τ and C_max were close to 100%, the study would have failed after scaling due to the upper cap at 50%. No α-ad­just­ment in a TSD? Strange. Or as nobody once wrote:

It’s originator, stupid!  Things are somewhat different in the non-generic world.