Therapeutic drug monitoring (TDM) of plasma samples is used in many circumstances to adjust pharmaceutics to desired concentrations. It is predominantly employed for drugs with a narrow therapeutic window (to avoid both underdosing and toxic levels) and/or where large inter- and intra-individual variations of concentrations are expected (e.g., antiepileptic drugs, anticoagulants, and immune modulators).

Antimicrobials: life-saving drugs with a difficult pharmacology

The effects of antimicrobials against pathogens are dependent on either peak concentrations or time above the minimal inhibitory concentration (MIC) or a combination of both, at the site of infection. Antimicrobial concentrations are influenced by drug distribution and elimination, which vary extensively in critically ill patients [1]. Therefore, treatment with “standard doses” of antimicrobials will be unpredictable [2]. Even though this knowledge is gaining momentum throughout the critical care community [3] the implementation of TDM has been slow [4].

TDM: why do we need it?

TDM of antimicrobials has the potential to detect and improve this variability [1]. While TDM was traditionally used to avoid specific toxicities of antimicrobials (e.g., glycopeptides or aminoglycosides), its focus has shifted towards improving therapeutic efficiency. β-Lactams are at the center of this development, but TDM of fluoroquinolones has recently gained more attention [4]. Apart from variations in target, the concept of TDM is applicable to most antimicrobials. The approach of controlling concentrations of antimicrobial drugs over time (pharmacokinetics, PK) to optimize their biological actions (pharmacodynamics, PD) is called “PK/PD-optimized dosing”. Several studies demonstrated the potential of TDM to improve the probability of target attainment (PTA) of antimicrobials in intensive care unit (ICU) patients. The use of TDM for PK/PD optimization has increased [5, 6], although technical demands (assay availability) and associated costs are still an obstacle. Beyond these barriers, other uncertainties need to be recognized.

Areas of uncertainty

What concentrations should be targeted and where?

There is significant heterogeneity in TDM studies regarding the definition of target concentrations, and the PTA of empirical dosing is dependent on if “high” or “low” targets are defined [7]. As an example, while there is consensus that concentrations for ß-lactams should be above the MIC of pathogens for the better part of a dosing interval, it is debated whether target concentrations of up to 4–6 × MIC add further benefit [1, 8]. As individual MICs are often not available, the use of empiric MICs for the respective antimicrobial, usually at the higher end of “sensitive”, is common. While the use of high MICs in combination with high target attainment aims at avoiding underdosing, the necessity of such high targets is questionable as it also increases the risk of attaining potentially toxic levels. It is unclear whether targets should be adjusted according to local resistance epidemiology and if TDM is particularly valuable in settings with high rates of resistance [9]. It must also be recognized that antimicrobial concentrations are usually measured in plasma samples, that not necessarily reflect concentrations at the site of infection and will thus act as a surrogate marker.

How do we adjust dosing when concentrations are out of range?

A recent study showed that the implementation of TDM in the ICU requires a dedicated effort and benefits from expert pharmacological advice. This might be an important component of antimicrobial stewardship programs in the ICU [10]. Unfortunately, there are no clear definitions of how doses should be adjusted in complex ICU patients. Many TDM studies assume a linear relationship [11], but situations in ICU patients are often dynamic. The two most important determinants of drug concentrations in critically ill patients, volume of distribution and clearance, may vary substantially, especially in patients with sepsis and septic shock. This is further complicated by extracorporeal organ support [12]. Various types of dosing calculation software are available, incorporating different pharmacokinetic models. Future models will likely be further enhanced with artificial intelligence. In such models, results from TDM are used to refine the modelling and enable ICU physicians to better individualize antimicrobial dosing [12].

Is steady state a prerequisite for TDM to offer meaningful information?

An often overlooked issue with TDM is the first 1–2 days of therapy. Most TDM protocols recommend sampling at steady-state level which is anticipated after 4 doses or at 24–48 h after initiation of therapy [13]. Even though this is meaningful for the correction of steady-state levels, it imposes a risk of not achieving antibiotic target concentrations during the “golden” first 24–48 h of therapy [14]. A loading dose or an extra dose within the first dosing interval is often recommended to more rapidly achieve target and steady-state levels. However, this is still empirical dosing and a protocol for earlier TDM, also in this early phase, could probably increase target attainment, but this will need further exploration Fig. 1.

Fig. 1
figure 1

Workflow for therapeutic drug monitoring with potential pitfalls

Full size image

What outcomes make sense when studying TDM in critically ill patients?

We do not know whether higher PTA of antimicrobials is beneficial to patients, and thus, the evidence in favor of TDM is still lacking. Recent trials did not show significant effects of TDM on mortality [4, 11]. However, a meta-analysis revealed positive effects of TDM on clinical and microbiological cure and treatment response [15]. Does this imply that we should not use TDM and halt its ongoing implementation? In an era of evidence-based medicine, this question can be answered with both yes and no. Since the evidence for benefit on patient-centered outcomes is still scarce, it could be argued that we should not introduce methods that are laborious and costly. On the other hand, current data suggest that TDM helps improve PTA by avoiding over- and underdosing. By gathering further data, different target achievements can be analyzed retrospectively and further enhance our knowledge. Also, besides patient outcomes, microbiological outcomes are important when discussing antimicrobial treatment. Although clinical data are lacking for this subject, adequate target attainment might reduce the emergence of resistant bacteria. Thus, there are also ecological benefits that need to be considered when analyzing the pro/con of TDM.

In conclusion, TDM will likely increase target attainment of antimicrobials in critically ill patients, reducing extremes of both low and high values with a possible positive effect on patient and microbiologic outcomes. There are still areas of uncertainty that will need to be further addressed in future studies.