The topic of software validation in pharmacokinetics is one of great debate now, and has been for many years. This debate often ends with absolutes like “you have to validate” or “there is no need to validate” the PK software being used. Software validation became a “hot” topic in the pharmacokinetic realm when the FDA released a guidance (or final rule) on 21 CFR Part 11 in 1997. However, the 1997 guidance was withdrawn and more flexible guidance documents were released in 2003 and 2007. You can find the gory details of the guidances here.
Instead of reviewing the guidances, I will give you my opinion of the validation required for PK software. Pharmacokinetics at its core is a complex set of mathematical equations. This means that anyone can take the concentration-time data and re-calculate the PK parameters if they have the proper equations. Thus, pharmacokinetics is no different from statistical summarization of data. The source data is always available for reanalysis at any time. This is distinctly different from a science like bioanalysis where a physical sample (e.g. blood, plasma, serum) is converted into a concentration measurement using physical means (e.g. LC-MS/MS). After the sample is exhausted, there is no way to perform another “analysis” to verify the concentration value determined previously.
With this background, it is my opinion that PK analysis software need only be validated to the extent necessary to ensure reasonable business confidence in the numbers generated by the software. I recommend a brief installation qualification followed by an operational/performance qualification step. Let me describe these 2 steps in more detail.
Installation qualification, or IQ, involves installing the software according to the vendor’s directions and making a record of that installation. This could be as simple as a single sheet of paper that is signed and dated by the individual performing the installation.
Operational/Performance qualification, which can be called OQ/PQ or OPQ, involves testing the software to make sure it performs as expected with a test dataset. This could be calculating a few PK parameters and comparing the output to previously calculated values using a different software package. Again, a sheet of paper indicating the testing that was conducted, and how the tested output compares with the expected output is generally adequate.
That’s it! Most validation experts and IT experts will try to tell you that more is needed. You can certainly do more, but I have passed an FDA inspection with no more than what was listed here. In the end, you are “validating” a calculator. And we all know that a calculator is only as good as the operator who inputs the numbers. Thus, more important that the software validation is the processes you have in place around your analysis to make sure that the calculated numbers make sense given the data used in the analysis.
So, I think you have to validate … but don’t go overboard. Keep it simple.
The Phoenix WinNonlin Validation Suite 7.0 is a desktop application that provides objective evidence of Phoenix WinNonlin 7.0 functionality and streamlines the validation process through the use of automated tests accessed through a graphical user interface and validation document templates. Read this brochure to learn more.