2013 - Glasgow - Scotland

PAGE 2013: Other Modelling Applications
Lia Liefaard

Predicting levels of pharmacological response in long-term patient trials based on short-term dosing PK and biomarker data from healthy subjects

Lia Liefaard (1), John Grundy (2), Nicola Williams (1), Andrew Shenker (3), Chao Chen (1)

(1) GlaxoSmithKline, UK (2) Isis Pharmaceuticals, Carlsbad, US (3) GlaxoSmithKline, US

Objectives: To develop a population PK/PD model using short-term dosing PK and biomarker data from healthy subjects; to apply this model to predict levels of pharmacological response, along with their uncertainty, in long-term treatment trials in patients.

Methods: PK and biomarker data from a short-term study in healthy subjects, in which a compound was administered at several dose levels, were analysed by means of population PK/PD modelling using NONMEM 7. Because of the expected long half life of the compound steady state was not reached in this study. The resulting PK/PD model was then used to predict the individual steady state biomarker levels in patients at the proposed clinical dose and administration frequency as well as sample size, taking into account inter-subject variability. For the simulations, the drug response was assumed to be generally comparable between healthy volunteers and patients, and the system part of the model was adapted for the patient population based on literature data. 300 replicates of the simulation were obtained, allowing for the uncertainty in the parameter estimates using the TNPRI subroutine of NONMEM [1]. The distribution of the proportion of patients reaching various specified levels of pharmacological response were calculated using the results of the 300 simulation replicates.

Results: A 2-compartment model with 1st-order absorption described the plasma PK adequately. The PK/PD model consisted of an indirect response model, in which the production of the biomarker is inhibited by the compound. An effect compartment accounted for the time differences between plasma PK and effect on biomarker levels. Using this model, including the parameter estimates and their uncertainty, the median and its 90% certainty of the proportion of patients meeting different levels of pharmacological response at steady-state could be predicted.

Conclusions: With this approach, the median proportion of patients in a long-term clinical trial meeting a desired level of pharmacological response and the certainty around this median can be predicted. In situations where certain levels of pharmacological response of a biomarker may be anticipated or known to correlate to a clinically meaningful effect, this approach allows for prediction of the probability of meeting such pharmacological response levels in a long-term clinical trial. This could then be used to justify dose selection or determine criteria for a futility analysis.

References:
[1] Boeckman AJ, Sheiner LB, Beal SL. NONMEM Users Guide - Part VIII. 1989-2011. Icon Development Solutions, Ellicott City, Maryland, USA.




Reference: PAGE 22 (2013) Abstr 2905 [www.page-meeting.org/?abstract=2905]
Poster: Other Modelling Applications
Top