I have been following the forum but this is my first post.
I am currently developing a sequential PK PD model for opioids in analgesia studies (I am working with veterinary species). The timecourse of antinociception is usually evaluated by applying a ramped thermal stimulus at regular time intervals. The PD dataset thus includes baseline temperature (before administration of analgesic) and threshold temperatures along time (onset to offset and return to baseline). I have data from several doses and administration routes to model together.
The testing device has a safety threshold (classically cut-off at 55C). At low doses, there is no problem, but at higher doses I have problems handling temperature data truncated by the safety threshold.
I have seen a couple of approaches to deal with truncated data (in NONMEM):
either using a time to event model as in Luks et al. 1998 (Weibull function) and Yassen et al. 2005 (Log logistic function) or
using an adaptation of the Beal M3 method to handle BLQ data, but this time with data above the limit of testing (as recently done by Sadiq et al. 2013)
I have tried a simple implementation of bql in the continuous observation block (bql = 55), but have the impression data above cut-off were just ignored. I have found some codes to implement M3 on NONMEM forums (see for example https://www.mail-archive.com/nmusers%40globomaxnm.com/msg03689.html) but I am struggling to translate them in a PML textual model that would run.
Does anybody have experience with either of these approaches for coding in Phoenix?
Many thanks in advance
Kind regards
Ludovic
Beal SL. Ways to fit a PK model with some data below the quantification limit. J Pharmacokinet Pharmacodyn. 2001 Oct;28(5):481-504.
Luks AM et al. Opioid-induced analgesia in neonatal dogs: pharmacodynamic differences between morphine and fentanyl. J Pharmacol Exp Ther. 1998 Jan;284(1):136-41.
Sadiq MW et al. Oxymorphone active uptake at the blood-brain barrier and population modeling of its pharmacokinetic-pharmacodynamic relationship. J Pharm Sci. 2013 Sep;102(9):3320-31. doi: 10.1002/jps.23492. Epub 2013 Mar 5.
Yassen A et al. Pharmacokinetic-pharmacodynamic modeling of the antinociceptive effect of buprenorphine and fentanyl in rats: role of receptor equilibration kinetics. J Pharmacol Exp Ther. 2005 Jun;313(3):1136-49. Epub 2005 Feb 8.
Please see the attached slides and the project attached that I believe will help you. We are talking about right censoring and how to define the likelihood.
welcome for your first post don’t hesitate to ask any question !
I had to use a pain inducing machine using an electric current in my PhD thesis and I am familiar with what kind of problems you might face in here.
Depending on the goal of your modeling you might need to consider a different type of analysis
But I would say ALQ would be good to try first as per serge post and let us know.
BLQ integrate up to the BLQ level you set so it is the reverse of what you need and it is natural that it did not work !
PML has special functions that facilitate writing the liklihood:
phi and lnorm.
note that there is cases where there is no real upper limit for response say infinity same when we integrate from - inf to BQL value and the code below show when we have a theoretical upper limit of your res