Sample size for HVD's [Power / Sample Size]
Dear Jay,
If you have to show bioequivalence also for the partial AUC's your sample size is clearly driven by the CV of that PK metric.
To go with a full replicate design has only advantages if you aim to show bioequivalence via scaled average bioequivalence. Sample size for the FDA recommended method for HVD's (target power 80%):
But that's in Europe a no go for AUC values. Here scaled ABE (via widened acceptance limits, ABEL) is only allowed for Cmax, which in your study came out as not highly variable.
Thus you have to stuck with conventional ABE.
See also this recent thread for a more extreme example.
❝ If CV of one parameter such as AUC5-t is observed more than 100% while for other parameter such as Cmax, Auc0-5 or AUCinf is within 20-30%. So should one go for full replicate design and sample size estimation based on highest CV observed as 114%.
If you have to show bioequivalence also for the partial AUC's your sample size is clearly driven by the CV of that PK metric.
To go with a full replicate design has only advantages if you aim to show bioequivalence via scaled average bioequivalence. Sample size for the FDA recommended method for HVD's (target power 80%):
library(PowerTOST)
sampleN.RSABE(CV=1.14, theta0=0.9, design="2x2x4")
++++++++ Reference scaled ABE crit. +++++++++
Sample size estimation
---------------------------------------------
Study design: 2x2x4 (full replicate)
log-transformed data (multiplicative model)
1e+05 studies for each step simulated.
alpha = 0.05, target power = 0.8
CVw(T) = 1.14; CVw(R) = 1.14
Null (true) ratio = 0.9
ABE limits / PE constraints = 0.8 ... 1.25
Regulatory settings: FDA
- CVswitch = 0.3
- Regulatory constant = 0.8925742
Sample size search
n power
44 0.79442
46 0.80099
But that's in Europe a no go for AUC values. Here scaled ABE (via widened acceptance limits, ABEL) is only allowed for Cmax, which in your study came out as not highly variable.
Thus you have to stuck with conventional ABE.
sampleN.TOST(CV=1.14, theta0=0.9, design="2x2x4")
+++++++++++ Equivalence test - TOST +++++++++++
Sample size estimation
-----------------------------------------------
Study design: 2x2x4 replicate crossover
log-transformed data (multiplicative model)
alpha = 0.05, target power = 0.8
BE margins = 0.8 ... 1.25
Null (true) ratio = 0.9, CV = 1.14
Sample size (total)
n power
372 0.800404
See also this recent thread for a more extreme example.
—
Regards,
Detlew
Regards,
Detlew
Complete thread:
- Sample size with CV more than 100% Jay 2015-11-02 10:47 [Power / Sample Size]
- Why not? d_labes 2015-11-02 11:08
- Why not? Jay 2015-11-02 11:29
- Sample size for HVD'sd_labes 2015-11-02 13:06
- EMA: MR – ABEL for partial AUCs Helmut 2015-11-02 14:35
- EMA: MR – ABEL for partial AUCs d_labes 2015-11-02 14:48
- EMA: MR – ABEL for partial AUCs Helmut 2015-11-02 14:35
- Sample size for HVD'sd_labes 2015-11-02 13:06
- Why not? Jay 2015-11-02 11:29
- Why not? d_labes 2015-11-02 11:08