EquivTest 2.0: defective! [Software]

posted by Helmut Homepage – Vienna, Austria, 2018-05-16 13:47 (1964 d 16:55 ago) – Posting: # 18769
Views: 5,259

Hi Irene,

❝ I conducted a data analysis for imbalance datasets from a journal entitled "Reference Dataset for 2-Treatment, 2-sequence, 2-period bioequivalence studies" (Dataset C) with EquivTest 2.0 and I compared the results with the ones stated in journal. I found that Equivtest 2.0 result (Point of estimate and 90% Confidence Interval) was not identical with EquivTest/PK (the result that stated in the journal). My Equivtest 2.0 result for point of estimates and confidence interval were:

❝ point of estimate (90% CI) : 66.78 (44.94,99.24)

❝ The result according to Journal (EquivTest/PK)

❝ point of estimate (90% CI) : 58.56 (39.41,87.03)

Congratulations! You discovered yet another defective software. :angry:
The result in EquivTest/PK agrees with ones of other software we have tested (SAS, Phoenix/WinNonlin, R). Screenshot:


The result you got in EquivTest 2.0 agrees with Kinetica 5.01 – which is wrong.1

Amazingly enough the correct formula taking the number of subjects / sequence (n1, n2) into account is given in the “User Reference Manual”, Chapter 8: Equivalence Testing (p. 166, p. 188 of the PDF) of v2.0 (dated 2001-10-12):


Did the developers update the manual but not the code

Anyhow, even if you upgrade to EquivTest/PK (of 2006)2 sooner or later you will face other problems. The Welch/Satterthwaite correction for parallel designs with unequal group sizes and/or unequal variances is not supported.3 Furthermore, you will not be able to assess replicate studies intended for reference-scaling according to regulatory requirements (FDA, EMA, WHO, ASEAN States, Australia, Brazil, Egypt, the Russian Federation, the Eurasian Economic Union, New Zealand).

I strongly suggest to get ‘better’ software.

  1. The defect for unbalanced 2×2×2 crossover designs was corrected in Kinetica 5.1 SR1 (2014-12-24). However, the flawed calculation of parallel designs with unequal group sizes was not corrected (see there).
  2. How? On the website of Statistical Solutions Ltd it is not listed any more.
  3. Fuglsang A, Schütz H, Labes D. Reference Datasets for Bioequivalence Trials in a Two-Group Parallel Design. AAPS J. 2015;17(2):400–4. doi:10.1208/s12248-014-9704-6. [image] free view-only version.

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

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

Complete thread:

UA Flag
 Admin contact
22,763 posts in 4,775 threads, 1,628 registered users;
8 visitors (0 registered, 8 guests [including 3 identified bots]).
Forum time: 06:42 CEST (Europe/Vienna)

The object of statistics is information.
The objective of statistics is the understanding of information
contained in data.    Irwin and Marylees Miller

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