Bioequivalence and Bioavailability Forum • Warning: BE in ThothPro

Warning: BE in ThothPro [Software]

posted by Helmut – Vienna, Austria, 2019-03-06 15:48 (1049 d 17:48 ago) – Posting: # 20000
Views: 3,743

Dear all,

I just tested the BE module of the current release 4.3.0 of ThothPro™.
I haven’t tried higher-order crossover and replicate designs yet.

Parallel designs are not implemented.
α can only be set in steps of 0.01 (though one has to specify 0.1 and not 0.05 to get the 90% CI). Too bad if you want 0.0294.
Unfortunate: The default is 0.01. If overlooked, one is punished with a 99% CI. Pretty conservative.
Funny: It is possible to select 1.00…
The lower and upper limits have to be specified separately (and not Δ like in PHX/WNL). That’s prone to errors. Furthermore, limits can be selected only in steps of 1%. Bad luck for people in Russia wanting to pre-specify 75.00–133.33% for C_max.
Outcome of our 2×2×2 reference datasets:
- The small balanced datasets (A, B, D, E, F) passed.
- The small imbalanced dataset C (n=13) failed with an outcome slightly different from infamous Kinetica.
  I got PE 66.78%, 90% CI 44.94–99.24%
  (correct: 58.56%, 39.41–87.03%; Kinetica: 66.78%, 44.91–99.31%). Would be interesting which formula is implemented.
- The large datasets G (balanced n=1,000) and H (imbalanced n=717) did not give any result at all.
  By trial-and-error: The maximum sample size is 100 (nothing stated in the manual).

Hence, I cannot recommend to use the software for BE till its glitches are fixed.

—
Dif-tor heh smusma 🖖
Helmut Schütz

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