Alan Abdulla

Critically ill patients represent a highly heterogeneous population with significant differences in the distribution of age, disease severity, length of stay, and clinical outcomes. Treatment in this population is therefore one of the most complex and expensive within healthcare. Moreover, infections in critically ill patients are a major source of morbidity and mortality in the intensive care unit (ICU). For the treatment of bacterial infections in critically ill patients, beta-lactams (penicillins, cephalosporins, carbapenems, and monobactams) and fluoroquinolones are the most commonly used classes of antibiotics. Rapidly achieving an adequate blood level of these antibiotics is associated with a greater chance of clinical success and a decrease in the risk of antimicrobial resistance. Because critically ill patients often have altered pharmacokinetics (PK) compared to non-critically ill patients, there is a significant risk that standard dosing regimens of these antibiotics can lead to sub-optimal exposure. As a result, there may also be variation between patients in the therapeutic response and the occurrence of adverse effects, making a ‘one-dose-fits-all’ approach undesirable. However, in patients with a severe infection the emphasis in practice is on the rapid initiation of therapy and the use of antibiotics with intrinsic activity against the causative pathogens. Unlike direct-acting drugs (e.g., inotropics, sedatives, and analgesics), where it is easier to titrate the dose to achieve a desired clinical response, antibiotics may take 24 to 72 h to show signs of adequate treatment of the infection. As a result, clinicians are generally less certain about an adequate dose selection to ensure good exposure.

To increase the likelihood of achieving good exposure with antibiotics, four main approaches can be used to modify standard regimens:

• dose nomograms for specific situations or populations;
• prolonged or continuous infusion, especially rational for antibiotics with a time-dependent effect (including beta-lactams);
• dose adjustment based on therapeutic drug monitoring (TDM);
• use of model-informed precision dosing (MIPD) in combination with TDM.

The latter two involve the use of population PK (popPK) models and captures antibiotic PK parameters, patient characteristics (e.g., age, sex and organ function), drug concentrations and disease characteristics (e.g., susceptibility to pathogens) in modeling approaches and then use Bayesian predictions to arrive at the optimal follow-up doses. This can reduce the variability in exposure.

In the ICU, optimizing interventions and therapeutic treatment is of great importance, but at the same time challenging and multifaceted. This is mainly due to the high degree of variability between patients and even within the hospitalization period of the individual patient. Optimization of the dosage of antibiotics in critically ill patients thus requires an individualized approach. To this end, this thesis describes advanced analytical methods to measure antibiotic concentrations in blood samples, characterizes antibiotic PK and its variability, assesses the impact of patient covariates relevant to clinical practice using nonlinear mixed effect modeling (NONMEM), followed by investigating therapeutic drug monitoring (TDM) and model-based dosing strategies. The studies in this thesis focus on the following commonly used beta-lactams and fluoroquinolones in the ICU: cefotaxime, ceftazidime, ceftriaxone, cefuroxime, amoxicillin (with or without clavulanic acid), flucloxacillin, piperacillin with tazobactam, meropenem, and ciprofloxacin. Based on the content of this thesis and its interpretation, we formulated specific recommendations for research, clinical practice, and policy.

Part I - Pharmacokinetic considerations in critically ill patients
In Part I, a general introduction is provided regarding the core principles of dose optimization in critically ill patients (Chapter 1). A good understanding of pharmacokinetics and pharmacodynamics (PK/PD) and the ability to properly describe them in specific populations is required to improve dosing in critically ill patients. The PK parameters describe the drug concentration over time after administration. These kinetic processes include the absorption, distribution, metabolism, and elimination of the drug. In particular, changes in the last three PK parameters should be considered important by clinicians when treating critically ill patients with intravenous antibiotics. Pharmacodynamics (PD) describes how biological processes in the body respond to or are influenced by a drug. The PK/PD relationship indicates the exposure-response relationship of a drug. For antibiotics, there are three types of relationships, namely time-dependent, concentration-dependent, and concentration-time-dependent. The most commonly used PK/PD indices of antibiotics are related to the lowest concentration of an antibiotic at which the growth of the bacteria is inhibited (minimum inhibitory concentration, MIC) and are divided into three indices:

• ƒT>MIC, the time (T) that the free concentration (ƒ) is above the MIC;
• ƒCmax/MIC, the ratio between the maximum free concentration (ƒCmax) and the MIC;
• ƒAUC/MIC, the ratio between the area under the free concentration-time curve (ƒAUC) and the MIC.

In Chapter 2, a literature review is provided of the relevant predictors associated with beta-lactam exposure during empirical treatment in critically ill patients. Rapid dynamic changes in physiological functions often underlie altered and variable PK parameters of antibiotics. However, adequate target attainment can be anticipated in critically ill patients prior to initiating empiric beta-lactam antibiotic therapy based on readily available demographic and clinical factors and augmented renal clearance are the most significant predictors for target non-attainment. For these strong predictors, we recommend their integration into practice. Identifying patients at risk is a first step in dose optimization and could help clinicians to consider more individualized dosing regimens and perform TDM when needed.

Part II - Analytic methods and population pharmacokinetic modeling
In Part II, analytical methods are described and population models are presented to study the PK parameters of antibiotics and the associated variability in critically ill patients. Chapter 3 describes an ultra-high performance liquid chromatography tandem mass spectrometry (UPLC-MS/MS) method using hydrophilic interaction liquid chromatography (HILIC). HILIC offers unique advantages for mass spectrometry detection of highly polar and ionic compounds (e.g., beta-lactams). This presented method was accurate, reproducible, with a short analysis time and was successfully applied in a PK/PD trial in the ICU (EXPAT trial). In addition, an UltraPerformance Convergence Chromatography (UPC2) method was developed. UPC2 is a more environmentally friendly chromatography method, as it requires fewer toxic solvents compared to normal or reverse phase chromatography. This novel, sensitive, and specific UPC2-MS/MS method demonstrated its value in the analysis of more than 800 human plasma samples in a clinical trial (DOLPHIN trial).

An understanding of the population PK in these patients and the ability to properly describe them is required to improve dosing in critically ill patients. With data from the EXPAT trial, we developed population PK models for cefotaxime (Chapter 5) and ciprofloxacin (Chapter 6) in a real-world cohort to assess the impact of patient covariates on the PK, using NONMEM. In Chapter 7, a pooled population analysis of ciprofloxacin is presented, using individual patient data from three trials. The PK differences between these trials were examined by a post-hoc analysis and the combined data were used to develop a pooled popPK model. The model was able to describe the population, but there was still unexplained interindividual variation. A simple ciprofloxacin dosing strategy suitable for all ICU patients remains a challenge and improved models or feedback of TDM may be needed.

Part III - Target attainment and therapeutic drug monitoring in clinical practice
In Part III, we describe target attainment and the use of TDM in critically ill patients. Empirical approach to dosages in critically ill patients results in failure to achieve desired targets exposure in a significant proportion of ICU patients. One way to achieve the target values is by applying TDM. By combining knowledge of PK and patient characteristics, TDM makes it possible to assess the efficacy and safety of a particular drug in different clinical settings.

In Chapter 8, we present the results of the EXPAT trial, the goals of this prospective study were to describe detailed target attainment of six frequently used beta-lactams in critically ill patients and to identify risk factors and clinical outcomes associated with target non-attainment. A total of 147 patients were included in the EXPAT trial, of whom 63.3% achieved pharmacodynamic targets of 100%ƒT>MIC and 36.7% achieved 100%ƒT>4×MIC. Regression analysis identified male gender, estimated glomerular filtration rate (eGFR) ≥90 ml/min/1.73 m2, and body mass index (BMI) as risk factors for target non-attainment. In patients who received continuous renal function replacement therapy or who had a high serum urea level, there was a significantly increased chance of reaching the target values. In addition, we found a significant association between 100%ƒT>MIC and ICU length of stay, but no significant correlation was found for the 30-day survival rate after ICU admission. These predictors, along with TDM, may help optimize beta-lactam doses in critically ill patients. Furthermore, we used data from the EXPAT trial to determine whether beta-lactam target attainment is a cost determinant in the ICU setting (Chapter 9). Renal replacement therapy was the most important cost driver and target attainment showed a trend toward higher total ICU costs (€44,600 vs. €28,200 per admission). However, this trend disappeared when correcting for ICU length of stay (€2,680 vs. €2,700 per day).

While the evidence for the added value of beta-lactam TDM is increasing, its clinical implementation remains limited. The aim of the review in Chapter 10 was to map the international use of beta-lactam TDM and to identify the opportunities and challenges of its implementation in critically ill patients. The main barriers were insufficient knowledge about various aspects related to the implementation and the unavailability of analytical methods. In addition, doubts about the cost-effectiveness of beta-lactam TDM in critically ill patients hinder broad implementation. In Chapter 11, we present the results of a nationwide cross-sectional online survey among Dutch healthcare professionals in which, among other things, barriers and facilitators for the implementation of beta-lactams and ciprofloxacin TDM have been identified. In particular, clear guidelines and organizational support are necessary for the implementation. Routine TDM requires preferably daily runs, simple sample preparation, while retaining sufficient assay sensitivity with quantification limits around MIC values of the most commonly causative pathogens. Finally, better evidence regarding favorable clinical outcomes from the use of beta-lactam and ciprofloxacin TDM may improve further implementation.

The effect of TDM on clinical outcomes, especially in critically ill patients, has been scarcely studied in randomized clinical trials. Moreover, recommendations on TDM are not unambiguous in the current guidelines. In Chapter 12, the DOLPHIN trial protocol is presented. In this multicenter, open-label, and randomized clinical trial, we enrolled 388 adult patients admitted to the ICU and treated with either a beta-lactam or ciprofloxacin. The aim of this recently completed trial was to assess whether the use of TDM and PK modeling can shorten the ICU length of stay (LOS) compared to standard therapy. The findings of the DOLPHIN trial will contribute to the knowledge and possibilities regarding optimization of the dosage of these antibiotics in critically ill patients.

In Chapter 13, we present a case report of ceftriaxone-related neurotoxicity in a critically ill child, and we discuss the current literature on ceftriaxone-related neurotoxicity with a particular focus on the role of the unbound ceftriaxone serum levels in critical illness. Common manifestations in critically ill patients, such as renal failure and hypoalbuminemia, appear to increase the likelihood of supratherapeutic ceftriaxone concentrations, which may contribute to an increased risk of developing ceftriaxone-related neurotoxicity. The availability of ceftriaxone analysis methods, both for total and free plasma levels, may play an important role in the prevention and diagnosis of ceftriaxone-related neurotoxicity.

Part IV - Future directions of therapeutic drug monitoring
In Part IV, we provide recommendations for clinical practice and future research on TDM. The rationale for performing TDM is the same as for other facets of precision medicine, which is to maximize clinical outcomes at the patient level. The position paper in Chapter 14 provides an overview of the current practice and future perspectives of antibiotic TDM in critically ill patients. By using TDM, under- and overdosing can be detected in time and thus contribute to an adequate antibiotic dosage. The use of TDM in critically ill patients may be further supported by MIPD. MIPD involves the use of population models and Bayesian predictions to reduce variability in response. A Bayesian approach provides estimated values for PK parameters, including the variability components, i.e. noise (residual error) and variability due to biological differences between individuals (inter-individual variability). In Chapter 15, we evaluated studies on the optimization of antibiotic dosing using MIPD in pediatric populations. In addition, we present a workflow for the implementation of MIPD. Although MIPD has the potential to improve the precision of antibiotic dosing, the broad integration of MIPD for antibiotics in pediatric clinical practice is still scarce. The studies we found were limited to vancomycin and amikacin. Available data on vancomycin indicate that MIPD is superior to conventional dosing strategies in achieving target values. To fully realize the potential benefits of MIPD, the tools must be implemented in a user-friendly framework for the healthcare team.

An important part of TDM is the blood sampling for the drug quantification. Sample collection for TDM is conventionally performed by venipuncture, which entails a specific blood draw and often the residual material is left unused. Given the global focus on sustainability, there is a growing shift from conventional blood collection to sustainable collection strategies. Accordingly, innovative and sustainable sampling strategies for the quantitative bioanalysis of drugs and metabolites have gained momentum. Several sustainable sampling approaches have been introduced to reduce the number of samples or total volume of collected blood, which can be categorized into three main groups: microsampling, sparse sampling, and scavenged sampling. Scavenged sampling can be a suitable approach for drug quantification to improve dosage regimens, perform pharmacokinetic studies, and explore the value of TDM without additional sample collection. The review in Chapter 16 elaborates on the current landscape of sustainable sampling within TDM and we provide a framework to stimulate sustainability, with a specific focus on scavenged sampling methods.

Finally, in Chapter 17 the main results of this thesis are discussed in a broader context. This chapter also sets out the vision for future research and implementation initiatives for TDM in critically ill patients, divided into three main themes: next-generation precision dosing, advanced matrices and biosensors, and sustainable sampling strategies.