## Adaptive Design for the FDA’s RSABE? [Two-Stage / GS Designs]

Dear all,

I stumbled across this (goody ?):

I don’t have the paper yet (behind a paywall) but I have some doubts that the method is well thought out:

I stumbled across this (goody ?):

**An adaptive trial design for testing the
bioequivalence of generics of highly
variable drugs**

^{1}a well-known framework^{2}is followed with Pocock’s \(\small{\alpha_\text{adj}=0.0294}\) (for*superiority*– the correct one for*equivalence*is \(\small{0.0304}\)…).I don’t have the paper yet (behind a paywall) but I have some doubts that the method is well thought out:

- \(\small{\alpha_\text{adj}=0.0294}\) in the two-stage approach for 2×2×2 designs (Method B) was a mere lucky punch and does not inflate the Type I Error
*only*for an assumed T/R-ratio of 0.95*and*target power of 80%. Any other combination requires other adjustments. Lots of papers…

Did the authors*believe*(‼) that Pocock’s \(\small{{\color{Red}{0.0294}}}\) is a ‘natural constant’ which is applicable in*any*adaptive design?

- Assuming a T/R-ratio of 0.95 for HVD(P)s is courageous. A more conservative 0.90 is recommended
^{3}and hence, also the default in the reference-scaling functions of the -package`PowerTOST`

.

- The authors’ -package
`adaptIVPT`

employs only 1,000 simulations by default (for T/R 0.95 and target power 80%). It is pretty fast because some parts are using`C++`

code and it supports multiple cores. Furthermore, \(\small{s_\text{wT}\neq s_\text{wR}}\) is supported.

RSABE itself is a framework (if \(\small{s_\text{wR}<0.294}\) → ABE; if \(\small{s_\text{wR}\geq0.294}\) → scaling*and*\(\small{PE\in\left\{0.8000-1.2500\right\}}\)). At least 100,00 simulations are required to obtain a stable result in the function`sampleN.RSABE()`

.

- Simulation-based adaptive designs require exploration of a reasonable narrow grid of stage 1 sample sizes and
*CV*for an assumed T/R-ratio and target power. Simulations have to be performed under the null hypothesis. Since the convergence under the null is poor, 10^{6}simulations have to be performed. Together with the 100,000 needed for the sample size, one would need 10^{11}simulations for every grid point.

- Under these conditions an \(\small{\alpha_\text{adj}}\) has to be found which maintains the empirical Type I Error.

- Since that was obviously not done, with \(\small{\alpha_\text{adj}=0.0294}\) the Type I Error possibly will be inflated. Furthermore, RSABE itself inflates the Type I Error if \(\small{s_\text{wR}<0.294}\), a fact the FDA prefers to ignore.
^{4}

BTW, inspecting the source code of`adaptIVPT`

I didn’t find 0.0294*anywhere*. Strange.

- IMHO, the method is questionable at least.

- Lim D, Rantou E, Kim J, Choi S, Choi NH, Grosser S.
*Adaptive designs for IVPT data with mixed scaled average bioequivalence.*Pharm Stat. 2023; 22(6): 1116–34. doi:10.1002/pst.2333.

- Potvin D, DiLiberti CE, Hauck WW, Parr AF, Schuirmann DJ, Smith RA.
*Sequential design approaches for bioequivalence studies with crossover designs.*Pharm Stat. 2008; 7(4): 245–62. doi:10.1002/pst.294. Open access.

- Tóthfalusi L, Endrényi L.
*Sample Sizes for Designing Bioequivalence Studies for Highly Variable Drugs.*J Pharm Pharmaceut Sci. 2012; 15(1): 73–84. doi:10.18433/J3Z88F. Open access.

- Schütz H, Labes D, Wolfsegger MJ.
*Critical Remarks on Reference-Scaled Average Bioequivalence.*J Pharm Pharmaceut Sci. 2022; 25: 285–96. doi:10.18433/jpps32892. Open access.

—

Helmut Schütz

The quality of responses received is directly proportional to the quality of the question asked. 🚮

Science Quotes

*Dif-tor heh smusma*🖖🏼 Довге життя Україна!_{}Helmut Schütz

The quality of responses received is directly proportional to the quality of the question asked. 🚮

Science Quotes

### Complete thread:

- Adaptive Design for the FDA’s RSABE?Helmut 2023-12-18 11:20 [Two-Stage / GS Designs]
- Likely it does not work (potentially inflated Type I Error) Helmut 2023-12-19 11:10
- Exploring package adaptIVPT, function rss() Helmut 2023-12-20 13:27
- Extreme test case Helmut 2023-12-24 13:01
- Extreme GMR Naksh 2023-12-25 04:16
- PE outside {0.80, 1.25} not possible Helmut 2023-12-25 10:54
- PE outside {0.80, 1.25} not possible Naksh 2023-12-25 11:42
- Forget rss() Helmut 2023-12-25 13:15
- Forget rss() Naksh 2023-12-26 04:49
- TSD useful at all? Helmut 2023-12-26 12:50

- Forget rss() Naksh 2023-12-26 04:49

- Forget rss() Helmut 2023-12-25 13:15

- PE outside {0.80, 1.25} not possible Naksh 2023-12-25 11:42

- PE outside {0.80, 1.25} not possible Helmut 2023-12-25 10:54

- Extreme GMR Naksh 2023-12-25 04:16

- Extreme test case Helmut 2023-12-24 13:01

- Exploring package adaptIVPT, function rss() Helmut 2023-12-20 13:27

- Likely it does not work (potentially inflated Type I Error) Helmut 2023-12-19 11:10