Pharmacometric methods in clinical pharmacology: An application to ketones

Abstract

Pharmacometric methods have seen increased application in clinical pharmacology in recent years. These methods are applied from the early stages of drug discovery to clinical practice. Pharmacometrics combines knowledge of mathematics and statistics to provide a framework to understand the time course of the pharmacological properties of drugs. This thesis is centred on development and application of pharmacometric methods, with an application to ketones as a motivating example.

Ketones are endogenous products of fatty acid metabolism which are produced in the liver in response to starvation. The three main constituents of endogenous ketones are D-β-hydroxybutyrate (BHB), acetoacetate (AcAc) and acetone. Of particular interest it is seen that elevated blood ketones (5 to 7 mM) are associated with therapeutic benefits in treating neurological disorders. Application of pharmacometric methods in this thesis is limited to BHB.

A population pharmacokinetic (PK) model was developed for BHB, in healthy adult volunteers following oral ingestion of a ketone monoester ((R)-3-hydroxybutyl (R)-3-hydroxybutyrate). This modelling process identified complicated input and disposition processes, including negative feedback on endogenous ketone synthesis. Potential covariates that influence the kinetics of ketones were identified in this work. This is the core work of this thesis that led to the subsequent work in rest of the thesis.

Issues with parameter estimation experienced during development of the PK model for BHB motivated the development of a new methodology for identifiability of population models. The method for identifiability developed in this thesis is an informal method, based on an information approach and use of the determinant of the Fisher information matrix. This is the first available method for assessing identifiability of fixed and random effects parameters in population models. Based on the inherent dependence of the information in a design and the associated random noise, this method can also assess deterministic identifiability of models for candidate designs. The method was evaluated using example population PK models. Application of this method was extended to assess the impact of parameterisation on identifiability of random effects parameters, using a one compartment PK model (Bateman model). This work highlighted that identifiability of random effects parameters in population models is sensitive to parameterisation and emphasised that care should be taken in parameterisation of population models.

The complicated input processes for ketones, identified in the population PK model for BHB, led to the development of a systems pharmacology model for catabolism of the ketone monoester. This systems model was explored for gastrointestinal absorption of ketones, in the form of in silico knockout variants similar to the knockouts that exist in in vitro and in vivo domains. The work identified that at least three distinct regions exists in the gut that contribute to the absorption of ketones. Exploration of mechanisms have indicated that different mechanisms dominate in different regions of the gut leading to spatio-temporal variation in the gastrointestinal absorption of ketones.

Finally, population models were developed for volume of oxygen intake (VO2) and the PK of BHB in healthy adult volunteers undertaking various intensities of physical exercise (workload). This work pertains to a small dataset that would help in the design of future studies with the ketone monoester. In this work, two independent population models were developed for VO2 and the PK of BHB. The model for VO2 is based on prior information from the literature related to workload and VO2. This model provided an acceptable description of the data. The model for BHB PK in this work was an extension of prior population PK model for BHB in this thesis. The effect of workload on the PK of BHB was introduced in the form of covariates on key parameters of disposition. This model performed reasonably well in describing the data. The modelling process has identified that the changes in VO2 and the kinetics of BHB are complicated at high intensity exercise and more work is needed to help describe this.

In conclusion, pharmacometric methods were successfully applied in understanding the pharmacology of ketones that would help us in designing future studies. The methodology for identifiability that was developed in this thesis was simple to use and can be used to (informally) assess structural identifiability of fixed and random effects parameters in population PK or PKPD models. Any optimal design software can be used for this purpose. The systems pharmacology model in this thesis can be adopted to explore similar compounds and the knockout concept can be used for mechanistic exploration of any compound.