jag009
★★★

NJ,
2016-06-04 00:39
(2854 d 11:41 ago)

Posting: # 16394
Views: 3,841
 

 Sample size for replicate: PowerTost & Endrenyi vs Davit [Power / Sample Size]

Hi everyone,

I was doing some RSABE refreshments (yeah getting old) and noticed a disparity between sample size generated by PowerTost & Endrenyi vs a paper from B. Davit et al.

Davit et al published a table showing the sample sizes for a 4-way replicate study design for a series of intrasubject CVs with GMR set at 1.0 and power=80%.

CV      Davit    PowerTost
30%      18      15
45%      34      16
60%      56      17
75%      80      19

I didn't tabulate results from Endrenyi's paper since the numbers are similar to the ones from PowerTOST. I can't confirm how Davit et al's paper arrived with their numbers since the article didn't mention the origin of those values.

The paper is:

B. Davit et al. Implementation of a Reference-Scaled Average Bioequivalence Approach for Highly Variable Generic Drug Products by the US Food and Drug Administration. The AAPS Journal, Vol. 14, No. 4, December 2012

Thanks

John
Helmut
★★★
avatar
Homepage
Vienna, Austria,
2016-06-04 03:34
(2854 d 08:45 ago)

@ jag009
Posting: # 16395
Views: 3,196
 

 ABE vs. RSABE

Hi John,

these sample sizes are for ABE; Barbara wrote above „As shown in Table I, the number of study subjects needed to show BE increases dramatically for HV drugs.”

CV%  GMR  Davit sampleN.TOST  EFG
30  1.00    18      16         16
    1.05    20      20         20
    1.10    36      34         34
45  1.00    34      34         34
    1.05    42      40         40
    1.10    72      72         72
60  1.00    56      54         54
    1.05    66      66         66
    1.10   118     118        118
75  1.00    80      78         78
    1.05    96      96         96
    1.10   172     170        170


In PowerTOST I got identical sample sizes with the exact method (Owen’s Q), the noncentral t-distribution, and the shifted t-distribution. Same sample sizes like PowerTOST’s in ElMaestro’s EFG. No idea how Barbara got her’s.

This is not what me worries in the paper. Since Detlew coined “FDA’s desired consumer risk model” hocus-pocus, this is the term László uses ever since. Once I had lunch with him and Barbara and she wasn’t able to explain what it actually means. We guess it’s sort of black magic to make to inflation of the TIE in RSABE look better. Let’s see her example on p.922:

library(PowerTOST)
sigma.0  <- se2CV(0.25)    # FDA's switching standard deviation
theta.s  <- log(1.25)/0.25 # FDA's regulatory constant
n        <- 36
design   <- "2x3x3"
CVwR     <- 0.30
sigma.wR <- CV2se(CVwR)
theta2   <- scABEL(CV=CVwR, regulator="FDA")[["upper"]]          # upper implied limit
TIE.1    <- power.RSABE(CV=CVwR, theta0=theta2, n=n, design=design,
                        details=FALSE, nsims=1e6)
theta2   <- ifelse(CVwR <= sigma.0, 1.25, exp(theta.s*sigma.wR)) # the hat-trick!
TIE.2    <- power.RSABE(CV=CVwR, theta0=theta2, n=n, design=design,
                        details=FALSE, nsims=1e6)
cat("TIE at the implied limits           :", TIE.1,
    "\nTIE of 'desired consumer risk model':", TIE.2, "\n")

TIE at the implied limits           : 0.132333
TIE of 'desired consumer risk model': 0.046205


She reported 0.046434. Hocus-pocus – the rabbit is back in the hat. Strong desires can move mountains.

See also:
Muñoz J, Alcaide D, Ocaña J. Consumer's risk in the EMA and FDA regulatory approaches for bioequivalence in highly variable drugs. Stat Med. 2016;35(12):1933–43. doi 10.1002/sim.6834

Dif-tor heh smusma 🖖🏼 Довге життя Україна! [image]
Helmut Schütz
[image]

The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes
UA Flag
Activity
 Admin contact
22,957 posts in 4,819 threads, 1,639 registered users;
70 visitors (0 registered, 70 guests [including 5 identified bots]).
Forum time: 11:20 CET (Europe/Vienna)

Nothing shows a lack of mathematical education more
than an overly precise calculation.    Carl Friedrich Gauß

The Bioequivalence and Bioavailability Forum is hosted by
BEBAC Ing. Helmut Schütz
HTML5