Where do these numbers come from? [Study As­sess­ment]

posted by Helmut Homepage – Vienna, Austria, 2015-11-23 15:01 (3367 d 08:54 ago) – Posting: # 15654
Views: 5,106

Hi Mahmoud,

Dear Sir

       Are you sure that Unique_One is male?


❝ 1. For the pilot fating study based on N=18.


Nope. n=16. :cool:

❝ The CV in cmax is >30% then we can use 3 way reference replicate scaled with N=36


Why 36? For FDA’s RSABE, CV 43.13%, T/R 0.90 (risky to assume a “better” one – even if the pilot suggests so), 80% power:

    design    n  power 
  RTRT|TRTR  18 0.82595
    RTR|TRT  28 0.83176
RRT|RTR|TRR  24 0.81442

The FDA requires a minimum sample size of 24, which will give you a power of 0.92498 if all subjects in the 4-period full replicate complete the study.

❝ or we can use two -stage 2x2 with N=50


How did you derive this number? A T/R of 0.9 calls for Montague’s “Method D” (an adjusted α of 0.028). Reference-scaling is not possible in TSDs. One can start the first stage with any number of sub­jects. A reasonable n1 is ~¾ of the fixed sample design’s n (with α 0.05: 154). With n1 116 power in the first stage is 59.8%, the chance to proceed to the second stage is 39.7%, and overall power is 82.6%. The expected average total sample size is 147. Percentiles of N:

min   5%  25%  med  75%  95%  max
116  116  116  116  184  222  310

If you start with only 50 subjects, power in the first stage will be just 28.4%, the chance to proceed to #2 is 71.6%, whereas overall power is similar with 81.4%. Generally you can expect higher sample sizes. Average 153 and:

min   5%  25%  med  75%  95%  max
 50   50   50  172  206  252  390

Not a good idea.

❝ 2. For fed study: The problem in GMR is about 87%. So in this case I recommened to genrate your study using the full replicate scaled with N=36


Again: Why 36? Most critical is AUCi with T/R 0.82 () and CV 35.48%. If we believe that these values are true (ha-ha), we would need 146 subjects in a 4-period full replicate. The 36 you suggested would give Unique_One a power of 45.7%. That’s less than the 48.6% winning chance one could expect for betting rouge/noir, pair/impair, manque/passe in roulette. Bad advice.
BTW, the high number is mainly caused by the [0.80–1.25] constraint on the PE. Without (theoretically) we would need 80 subjects.

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

Complete thread:

UA Flag
Activity
 Admin contact
23,379 posts in 4,913 threads, 1,661 registered users;
107 visitors (0 registered, 107 guests [including 17 identified bots]).
Forum time: 23:56 CET (Europe/Vienna)

Science is what you know.
Philosophy is what you don’t know.    Bertrand Russell

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