Bioequivalence and Bioavailability Forum

Dear Venu!

❝ Is it compulsory to use WinNonlin software for generating pharmacokinetic and statistical analysis for pilot and pivotal studies?

No!
Actually there is no regulation forcing you to use any software at all…
You may also calculate your ANOVA by means of (some) sheets of paper, a pencil and a pocket calculator (although in order to avoid errors, you should repeat this procedure, or let two persons do the job).

❝ If not what are the other software alternatives?

Kinetica from ‘Thermo Electron’ (not only BA/BE, but also NCA, ‘classical’ PK and population PK).
Drawbacks (up to the current version 4.4.1):

• Besides an acceptance range of 0.80–1.25 only 0.70–1.43 is supported… (what about 0.75–1.33, 0.90–1.11)?!
  
• Only 90% confidence interval possible (what about Brazil and multiple testing)?

EquivTest/PK from ‘statistical solutions’:
Drawbacks:

• Only 2×2 designs are covered.
  
• Acceptance ranges only 0.80–1.25 and 0.70–1.43!

Drawbacks of WinNonlin and Kinetica: both pieces of software are lacking the nonparametric method (which you need in many regulations for t_max); therefore you will need EquivTest anyhow.

An alternative for nonparametrics is StatXact from ‘Cytel’, which also comes as a plugin (StatXact PROCs) for SAS.

If you have a large budget – and the additional salary for a SAS-programmer - you also may use SAS for the entire evaluation…

If you know a statistician with some programming skills, you may adapt routines in any general-purpose statistical package (e.g., Stata, NCSS, Statistica, SPSS, Minitab, Systat, etc).

If you employ a rather skilled programmer (steep learning curve!), you may also start from scratch either in S-Plus from ‘Insightful’, or in the – equally powerful open source software – R by “The R Project for Statistical Computing”.
If you are interested in PK, you also may go for David Bourne’s Boomer (for Macs, Windows, Linux) and the new graphical interface JGuiB.
A (Bayesian) statistical freeware is WinBugs (with the plugin for PK-analysis PKBUGS).

❝ Is it also compulsory to validate the WinNonlin software?

Not only WinNonlin, but all software…
I would not recommend buying one of the pretty expensive “validation packages” offered by the software vendor (some of them are more expensive than the software itself), but applying the classical validation approach:

• lay down a written validation plan (what will be done – for which purpose – how often<– by whom…)
  
• get example data sets (from standard textbooks, NIST, …)
  
• set specifications (e.g., maximum deviation of calculated point estimates and confidence limits ±0.01% from reported ones)
  
• write an SOP
  
• perform the validation according to the SOP
  
• revalidate the software every time you update the software and/or the operating system