Abstract by Theodoros Papathanasiou

The development of new drug treatments is a long, risky and costly venture. This is even more so the case when developing drug combinations, where a multitude of potential dose levels can be considered during the development process. Despite the long history of applying model-based approaches for the analysis and interpretation of pharmacodynamic drug interactions at the preclinical level, there is a substantial knowledge gap when it comes to the feasibility of applying such methods to clinical data with inherent large variability in their measured drug response, such as body weight change following antiobesity pharmacotherapy. Thus, the aim of this PhD thesis was to develop and apply pharmacometric methods to optimize the dose-finding stage of the drug development process, with a special focus on drug combinations, and the pharmacological treatment of overweight and obesity.

The feasibility of Dose-Exposure-Response (DER) analyses for dose-finding studies in drug combinations was explored via the means of clinical trial simulation. Dose-finding studies for a hypothetical glucagon-like peptide-1 (GLP-1) receptor agonist combined with a novel add-on for the management of overweight and obesity were used as a case study. The underlying DER relationship of the hypothetical GLP-1 receptor agonist was inspired by the previously delineated DER relationship of liraglutide. Four realistic DER relationships were evaluated under eight potential dose-finding designs, to evaluate the reliability of parameter estimates and the probability for accurate dose identification. Overall, the study demonstrated that pharmacodynamic interactions models can be used for the DER analysis of clinical endpoints especially when information regarding the DER profile of one monocomponent is available. Additionally, the study highlights that at least one dose of both monocomponents should be administered as monotherapy in order to obtain an accurate delineation of the underlying DER relationship.

The added value of using model-based optimal design (OD) for guiding the dose allocation in drug combination dose-finding studies as compared to a typical drug-combination trial was explored. For comparability, the same DER relationships as those explored during the feasibility evaluation were used. A novel method for defining the optimal minimum combination dose for a selected target-effect level was combined with a compound optimality criterion, which assured benefits for both parameter estimation and precision of model predictions. Optimal designs were found to lead to unbiased estimates and significant improvements in the accuracy of results relative to the typical design, thus demonstrating that the optimal design methodology in tandem with DER analyses is a beneficial tool that can be used for appropriate dose allocation in dose-finding studies for drug combinations.

To characterize the weight-loss time course following treatment initiation with the GLP-1 receptor agonist, liraglutide, a novel population pharmacokinetic/pharmacodynamic model was developed. The developed model was able to reveal two phases in the weight loss trajectories, a transient and a more sustained one, as well as a subtle signal of seasonal variation in weight change. The model was used to explore slower treatment initiation escalation algorithms than the one described in the prescribing information for liraglutide 3.0 mg for weight loss. The model simulations suggested that slower dose escalations at treatment initiation can be applied with only a modest impact on the expected weight loss outcome. These results are potentially valuable for patients who experience gastrointestinal side effects and would benefit from a slower dose escalation.

The methodology of model-based adaptive optimal design for nonlinear mixed-effects models was extended to include a model selection (MS) or a model averaging (MA) step during the interim evaluations in order to account for model structure uncertainty. The extended model-based adaptive optimal design methodology was applied to a dose finding study of a hypothetical GLP-1 compound intended for the management of overweight or obesity. Response adaptive allocation probabilities were updated at each interim evaluation step and were driven by the D-optimality criterion, computed for the simulation model (SM), the best model in terms of Akaike Information Criterion (AIC) (for MS), or all weighted models (for MA). A stopping rule was included to stop the trials early for futility. The designs were compared using the efficiency relative to the D-optimal design, and dose-finding predictive performance criteria. Model-based adaptive optimal designs with MS or MA were shown to lead to similar performance as compared to the best-case situation, where the correct model structure is assumed to be known a priori. These results indicate that the assumptions regarding the model structure can be relaxed with minor impact on the adaptive process, thus making model-based adaptive optimal design less model-dependent and potentially more attractive for increasing the efficiency of dose-finding programs.

In conclusion, pharmacometric models and methods were developed and applied to improve the design and analysis of dose-finding clinical trials, with a special focus on drug combinations and the pharmacological treatment of overweight and obesity.