The 6 biggest pharmacometrics modelling mistakes!
Alan Maloney
Equation AB
Objectives:
Applied pharmacometrics is very difficult. From ensuring the designs of clinical studies are sound, through to analyses using non-linear mixed effect modelling, there are many opportunities for things to go wrong.
The objective of this presentation will be to highlight a number of the main technical mistakes I have made (or have seen) in applied pharmacometrics, and offer potential solutions.
Whilst I am hugely indebted to my educators, a number of mistakes can be traced back to my early education, where the desire of my tutors to provide a concise framework for modelling resulted in subtle, but important, factors being omitted. The “cookbooks” are not always right.
Methods:
The 6 mistakes that will be covered:
- Mistakes in experiment design (complex models versus weak data)
- Mistakes in model selection (significance testing and parsimony)
- Mistakes in model assessment (how sound are DV v PRED/IPRED?)
- Mistakes in model qualification (useless predictive checks)
- Mistakes in model uncertainty (or why didn’t we learn this!)
- Mistakes in parameter uncertainty (say no to the NP bootstrap!)
Advice or alternative methods and approaches will be suggested.
Conclusions:
The presentation will hopefully be enlightening, and help stimulate discussion on improving how we design, execute and evaluate our pharmacometric projects. It is also hoped the audience will be inspired to challenge their own education “cookbooks”.
