Use of VIF

I modelled a pooled plasma concentration-time profile in individuals and then simulated a profile using blood sampling times as per protocol. I then excluded sampling times (up to 3 times) and looked at the ratio of VIF (min sampling/max sampling) at each sampling time. I took a ratio of <1.2 as acceptable (this is the subjective step). I could then predict that if a subject missed, in this case, 3 samples, PK parameters would still be reliably estimated. This approach was based on: Gabrielsson and Weiner (2006) Pharmacokinetic and Pharmacodynamic Data Analysis: Concepts and Applications, 4th ed. Swedish Pharmaceutical Press, Stockholm.

I understand that I could also compare VIF for each parameter of interest.

My question: is this a reasonable approach? Is a ratio of 1.2 acceptable?

think of this ratio is how much precision percent you are losing if degenerating your precision of parameter estiamte by 20 % is acceptable then < 1.2 can be fine

This makes sense,

Thanks, Charlie

Another way I’ve used this is to try different times from the original study e.g. I have modelled that data and found I think Cmax occurs at e.g. 1.28 h so maybe I can improve my chance of capturing Cmax in future studies by trying different sampling schedules;

0, 0.5, 1, 1.5… original

0,0.66,1, 1.33… test 1

0, 1, 1.25, 1.5 … test 2