Daniel Lill (1,3), Anne Kümmel (1), Nathalie Gobeau (2), Henning Schmidt (1), Jens Timmer (3)
1) IntiQuan GmbH, 2) Medicines for Malaria Venture, 3) University of Freiburg
Introduction
Dose selection in clinical trials is often based on the probability that the chosen dose achieves the desired clinical outcome. Simulation studies assessing a range of doses and accounting for inter-individual variability as well as parameter uncertainty are leveraged to solve this implicit problem. This procedure is however impeded for combination therapies due to combinatorial complexity: Compared to one optimal dose for monotherapy, many dose combinations, i.e., isoboles, lead to the same desired effect. The computational effort determining isoboles taking into account IIV and uncertainty by a brute force approach of screening the whole higher dimensional dose grid for these isoboles becomes infeasible. However, simulations of only the typical behavior (neglecting IIV) may not necessarily represent the population behavior for non-linear models. Therefore, we extend the use of isoboles to population simulation by an efficient new algorithm and show how statistics can be performed in the two-dimensional combination therapy setting. The algorithm’s performance and the use of population isoboles are exemplified by three different drug-drug-interaction PKPD models from Malaria research.
Methods
A new algorithm to calculate isoboles was developed to reduce the computational effort of isobole calculations to allow their use in population simulation. Instead of simulating the model for all possible dose combinations in a given dosing range, the algorithm iteratively refines the resolution of the dosing grid close to the actual isobole. It is robust with respect to the shape of the isoboles and shows good convergence properties even for strongly non-linear models. To obtain statistical properties such as the median isobole and the 95% confidence intervals, an additional algorithm for two-dimensional quantile calculation was developed.
Results
We present the properties of the algorithm and features of population isoboles by comparison of three different models from malaria research. For each model, the dose combinations yielding a 95% cure rate in the population, the target for antimalarial drug treatment, are determined. The isobole algorithm cuts down the required simulations by more than a hundred fold, allowing to unite the approaches of population simulations and combination therapy modeling. The simulation results show that population isoboles can be remarkably different from single subject isoboles, emphasizing the importance to consider population effects. Furthermore, the newly developed method to obtain confidence intervals for isoboles is a powerful yet simple way to quantify model uncertainty for combination therapies.
Conclusions
Population isoboles provide a visual as well as quantitative means to assess the model behaviour and uncertainty across the possible dosings. Compared to isoboles of a single subject, they are better able to capture extremal behaviours in the population. Taken together, population isoboles provide an efficient tool to identify interesting dosing regimens and are easily communicated to collaborators. The computational efficiency of the new algorithm allows its application in a study comparing several anti-malarial combination therapies to select the most promising one to be tested in a clinical trial.
References:
[1] An Overview of Drug Combination Analysis with Isobolograms, Ronald J. Tallarida, Journal of Pharmacology and Experimental Therapeutics October 1, 2006, 319 (1) 1-7; DOI: https://doi.org/10.1124/jpet.106.104117
Reference: PAGE () Abstr 9540 [www.page-meeting.org/?abstract=9540]
Poster: Oral: Methodology - New Tools