Brendel
SERVIER
Objectives: A change-point model is a model in which a parameter's value discontinuously changes at a given time (named change-point). It can be used to modify model behaviour (e.g. shift in absorption parameter value) or update the system (e.g. Enterohepatic recycling). In addition to parameter values, such models require the definition of change-point, and its implementation in NONMEM is not as intuitive as it looks, especially in repeated doses and at steady state.
Methods: Three methods of change-point definition were evaluated: firstly the reference method, using model event time parameter (MTIME) directly implemented in NONMEM (MTIME_meth); secondly the use of a fictive change point compartment (CMT) with an associated lag time (CMT_meth) [1]; finally a change point function (FCT) with an analytical expression of a rectangular wave function (FCT_meth).
Each method was evaluated with a simulated population PK example: a 2 compartment model with a changing first-order absorption rate. Simulations were performed after single administration, repeated doses and at steady state and 90% prediction interval (PI) and the median were plotted. Graphical comparison and run time duration were used to evaluate each method. Practical constraints in terms of coding and/or data preparation were also discussed.
Results: After single administration, the three methods give the same expected results. After multiple administrations, no difference in term of simulation output is observed between the three methods. As the reference method MTIME_meth allows the use of an adapted NONMEM subroutine, run duration is smaller than CMT_meth and FCT_meth, which require using differential equations. At steady state (using the implemented NONMEM SS option) only the FCT_meth is able to give the correct simulation results. From a practical point of view, MTIME_meth involves heavy code adjustment and CMT_meth requires specific dataset modification, while FCT_meth can be used without significant modifications of both code and dataset.
Conclusions: In order to implement a change-point model after a single dose administration, the MTIME_meth was the easiest way to do it, allowing to the modeller to use NONMEM subroutines.For multiple dosing, the CMT_met and FCT_met are good alternatives when the number of doses increases. However, only the FCT_meth is able to perform correct simulation at steady state using NONMEM SS option.
References:
[1] EJPAGO 19 (2002) Abstr x. Owen
Reference: PAGE 22 () Abstr 2912 [www.page-meeting.org/?abstract=2912]
Poster: New Modelling Approaches