I just got a request from a way higher level than I am, so the (as I think) correct answer of “that is not the way it is intended to be used” is not accepted.
I provided a simple NPS as an estimate of a steady state exposure for a compound with a very low t1/2. That’s OK for me, no accumulation resulted in an easy QC of the results.
To be on the safe side, I now shall provide another superposition, but this time the maximum t1/2 seen in the population shall be applied to each subject.
However, as far as I can see, such a scenario is not possible using the settings of the NPS available in Phoenix WNL, or is it?
That is, as a work-around I would personally simply calculate one (or two, if the N not less than three rule applies here) additional concentrations after the real Clast, append them to the data and force WNL to use these time points for the Lz-estimation during NPS, but maybe there is an easier way to do this?
Any comments or suggestions will be greatly appreciated.
Yes you’re right, if you want to use NPS then you’ll need to supply 2 points as Start and End to the terminal phase to the end of your profile. Probably you can use the Predicted conc from the last 2 points in Summary Table for the NCA of the profile with this most ‘extreme’ Lz and apply to all profiles. (NPS will accept just 2 points)
Or you could simulate using modelling to give you these 2 points.
But if you’re doing that why not do perform simulations instead of NPS? It’s easy enough to add a couple more ‘profiles’ of e.g. 50% CL, 150% CL etc.
Non-parametric superposition assumes that the profile is representative and it will “apply” the terminal phase of the profile to do the superposition. You can tweak the terminal phase but how valid it is to only tinker with a part of the profile ?
I would suggest to conduct a model based simulation because your are modifying the assumptions/parameters that were “observed”.
Well, you may know the most relevant statement in every clinical trial “But it is stated in the protocol”. Yeah, tinkering with the data is not stated in the protocol either. But it is not forbidden and the documentation will be state of the art. That much for simulation .
But beside such formalities: I think they wanted to avoid anything that might result in discussion. That is, modelling and simulation will always depend on at least some input from the responsible person choosing the model, maybe defining starting values etc. That input needs to be communicated or pre-set by SOPs, both need to be understood and someone may will start a discussion.
Superposition is clean and easy and simple to understand. And everyone is quite familiar with descriptive statistics on PK parameters. So that is a low-hanging fruit.
I do not fully support such thinking but I understand the approach.
I do fully agree with you, Samer, that using Lz from one subject in order to “increase” the accumulation in another subject is as logical as harvesting the subject’s kidneys would be. Admittedly, both ways we artificially decrease the clearance.
But, to alter the observed parameters (either in a model or by abusing Lz of someone else) will just add a theoretical exposure in someone who does not exist but happens to have a lower clearance.
I fail to understand the whole purpose of such an exercise, because I do not need any numbers to predict, that exposure in general and in particular during the dosing interval will increase with a decrease in clearance/when I start to see accumulation.
However, from a technical point of view I now know what to propose (I don’t think I can use one set of concentrations from the “worst” subject, Simon. I think, also these concentrations woukld then be used in the superpositioning itself and not only in the estimation of Lz.). Adding one single additional concentration to each data set sounds easy enough.
Thanks for the highly appreciated input and best regards,
.