Bioequivalence and Bioavailability Forum

Hi Osama,

❝ I'm not sure but I think this is the first time […] in which the WHO mentioned to the two-stage sequential design in it's BE guidelines

Correct.

❝ […] WHO described it generally and didn't mentioned to any certain method (potvin B or C), Montague (method D) and which one they recommend.

Unfortunately WHO followed in the wording EMA’s approach. My interpretation of lines 1237–44

In order to employ a two-stage design, adjustments must be made to protect the overall Type I error rate and maintain it at 5%. In order to do this both the interim and final analy­ses must be con­ducted at adjusted levels of significance, with the confidence intervals calculated using the ad­justed values.
It is recommended that the same alpha for both stages be employed which gives an alpha of 0.0294 for this case,

speaks for Method B (α 0.0294) because in the first stage of Method C α is only adjusted if power <80%. The text continues with

However, the amount of alpha to be spent at the time of the interim analysis can be set at the study designer’s discretion. For example, the first stage may be planned as an analysis where no alpha is spent in the interim analysis since the objective of the interim analysis is to obtain in­for­mation on the point estimate difference and variability and where all the alpha is spent in the final analysis with the conventional 90% confidence interval. In this case no test against the acceptance criteria is made during the interim analysis and bio­equi­valence cannot be proven at that point.

I’m not aware of any publication exploring no (!) adjustment in the first stage and stopping for futility. O’Brien-Flemming with α (0.005, 0.048), expected GMR 0.95, target power 80%, CV 20%, stage 1 sample size 12, and a futility criterion of 0.85 ≤ PE ≤ 1/0.85 leads to a slight inflation of the Type I Error. Trying

library(Power2Stage)
power.2stage.fC(alpha=c(0.005, 0.048), CV=0.2, n1=12, GMR=0.95,
  targetpower=0.8, fCrit="PE", fClower=0.85, theta0=1.25, nsims=1e6)

gives

TSD with 2x2 crossover
Method B: alpha (s1/s2) = 0.005 0.048
Interim power monitoring step included
Target power in power monitoring and sample size est. = 0.8
Power calculation via non-central t approx.
CV1 and GMR = 0.95 in sample size est. used
Futility criterion PE outside 0.85 ... 1.176471
BE acceptance range = 0.8 ... 1.25

CV = 0.2; n(stage 1) = 12; GMR= 0.95

1e+06 sims at theta0 = 1.25 (p(BE)='alpha').
p(BE)    = 0.053862
p(BE) s1 = 0.021617
Studies in stage 2 = 18.31%

Distribution of n(total)
- mean (range) = 13.8 (12 ... 78)
- percentiles
 5% 50% 95%
 12  12  24

❝ does this mean they will accept all?

WHO’s guideline are intended to support regulators of countries which don’t have their own (local) guide­lines in place. I guess if one wants to use anything deviating from variants of Method B he/she has to show in simulations that the overall-α is maintained.