Use of Physiologically-Based Pharmacokinetic Modelling To Support Dose Optimization in The Critically Ill

Abstract

Critical care is a subspeciality of medicine that incorporates a multidisciplinary approach for patient care in a heterogeneous population. The wide array of etiologies to critical illness may result in disturbances to homeostasis and organ function. This variability in host-response may impact the pharmacokinetics (PK) of an individual, resulting in potential differences in drug exposure compared to a healthy counterpart. The high mortality rates of critically ill patients are thought in part to be due to altered drug exposure, resulting in suboptimal dosing or adverse effects. The complexity and heterogeneity of critical illness limits the feasibility of conventional PK studies for dose determination in this population. Furthermore, current “one-dose-fits-all” approaches in pharmacotherapy overlook the interplay between drug physicochemical properties and pathophysiology towards altered PK in the critically ill. Methods incorporating pharmacometrics have advanced our understanding of this relationship, however current approaches may require robust data sets to inform dose-exposure relationships and limit extrapolations to untested clinical scenarios. Methods incorporating physiologically-based pharmacokinetic (PBPK) modelling are encouraging to overcome some of the current challenges and knowledge gaps towards optimizing drug therapy in critically ill patients. PBPK models leverage mechanistic principles to incorporate knowledge of physiology and drug physicochemical properties to provide a simulation-based approach to drug PK. Knowledge of drug PK and physiology in healthy adults may be adjusted to account for differences in age (i.e adults to children), or disease (i.e. critical illness), to predict drug exposure. As these methods leverage known prior information, it enables a more proactive approach in predicting drug PK in clinical settings when data sampling is sparse, such as the critically ill. This thesis aims to apply state-of-the-art mechanistic modelling strategies to advance pharmacotherapy in critically ill adults and in children receiving extracorporeal life-saving technology. The objectives are to (1) analyze the PK data in critically ill children receiving continuous renal replacement therapy (CRRT), (2) evaluate the extraction of drugs via CRRT utilizing a closed-loop ex-vivo study design, and (3) develop PBPK models to optimize drug dosing in various presentations of critical illness. The first objective included a systematic review to evaluate our current understanding of drug PK in critically ill children receiving CRRT. Several knowledge gaps were identified, which were to be addressed within the second objective by incorporating a closed-loop study design to isolate the drug-CRRT circuit interaction. This ex-vivo¬ approach was the first of the holistic methodological solution towards which this thesis aims to achieve. Knowledge gained from the ex-vivo studies may be integrated into a PBPK model structure to provide a mechanistic understanding of drug PK, allowing for prospective dose predictions in untested scenarios. Within this work, it was identified that 40% of PK studies conducted in critically ill children receiving CRRT focused on antimicrobials. Furthermore, only 50% of studies included in the final analysis provided dosing guidance based upon PK findings in this population, representing several knowledge gaps to support clinicians in dosing guidance at the bedside. Anakinra and cefazolin were targeted as drugs of need for dosing guidance in patients receiving CRRT. Ex-vivo studies were conducted to investigate the drug-circuit interaction. Both anakinra and cefazolin, were efficiently removed from plasma upon the completion of the experiment, rendering a sieving coefficient of 0.34 and 0.31, respectively. A PBPK model for anakinra was developed to predict the PK in children receiving concurrent CRRT and extracorporeal membrane oxygenation (ECMO). Upon applying the results from the ex-vivo experiment, PBPK model predictions successfully described anakinra PK amongst critically ill children enrolled in a prospective, opportunistic study. Overall anakinra exposure was similar amongst children receiving ECLS compared to those who are not, supporting the hypothesis that renal dose adjustments in this population may not be required. The work herein this thesis seeks to advance the application of PBPK in supporting pharmacotherapy to a diverse group of critically ill patients. In doing so, a mechanistic representation of cefazolin drug concentrations in the central nervous system (CNS) through PBPK was performed. Cefazolin cerebrospinal fluid (CSF) concentrations were accurately predicted, and simulated CSF:plasma ratios fell within a 1.5 fold-error compared to observed values. Dosing simulations demonstrate that a continuous infusion of 10 g/day may be required to achieve pharmacodynamic response towards methicillin-susceptible Staphylococcus aureus in treating infections in the CNS. This thesis aims to articulate the multiplicity of PBPK to support dosing guidance in the critical care setting. Several knowledge gaps were identified, and the methods applied aim to address some of the current limitations of conventional PK studies in this population. PBPK modelling efforts such as these have the potential to improve generalizability, and efficiently conduct PK studies in a highly vulnerable patient population by reducing the number of participants required and applying a more proactive approach to dose determination in untested clinical scenarios.