Critically ill patients commonly present significant pathophysiological changes that modify the PK of antimicrobials [20]. Particularly in septic shock, the blood flow of the gastrointestinal tract and subcutaneous tissue may be severely reduced and shunted to vital organs [21] such as the brain and heart. This compromises reliable drug absorption through these routes. As a result, the intravenous administration of antimicrobials is commonly recommended [22].

These changes, noted in sepsis, result from a significant fluid shift from the intravascular compartment to the interstitial space [23]. These shifts are the product of inflammatory endothelial damage and increased capillary leakage [24], which can be aggravated by aggressive intravenous volume replacement. Hypotension and hypoperfusion can also damage organs and tissues, further changing the antibiotic concentration and PK.

The antibiotics’ volume of distribution (Vd) is often increased [21,25], and consequently, as β-lactams are hydrophilic antimicrobials, there is a need for higher loading doses to achieve the same adequate therapeutic concentrations.

The β-lactams’ Vd is also influenced by protein-binding modifications [26], particularly albumin, one of the main plasma-binding proteins for many drugs [27]. Septic patients commonly have decreased concentrations of albumin, which directly impacts the PK of antibiotics. Hypoalbuminemia increases the unbound antibiotic concentration [26], which leads to a lower, probably suboptimal, antimicrobial concentration. These should be accounted for if therapeutic drug monitoring (TDM) is used, and the direct measurement of free drug levels should be preferred [28,29].

Not only do these pathophysiological changes directly influence antibiotic concentrations, but therapeutic interventions can also cause dramatic modifications of the antibiotic PK [30]. Invasive mechanical ventilation, abdominal surgery, and high-dose vasopressors are among the interventions that contribute to fluid shifts and changes in drug concentration.

Acute kidney injury and a reduced glomerular filtration rate [28] are also common complications of sepsis. Accordingly, decreased drug clearance (and a longer half-life) may occur, making antibiotic PK and concentrations largely unpredictable [31]. In addition, the evaluation of renal clearance may be quite challenging due to frequent changes in the fluid balance and the poor predictive value of serum creatinine [31].

Even in patients without acute kidney injury, the antibiotics’ serum concentrations may be severally altered due to augmented renal clearance [20,31]. This is the case of antibiotics predominantly excreted by the kidney, such as β-lactams. During augmented renal clearance, β-lactam elimination is increased, and the consequent subtherapeutic concentrations can jeopardize the efficacy of antibiotic therapy [32].

These PK changes are dynamic, and daily changes are very common, leading to the frequent need to adapt the antibiotic dosage. Antibiotic concentrations can be quite unpredictable, and failure or toxicity, which are directly related to a change in the concentration, may become frequent [21].

Furthermore, the relationship between the amount of drug being administered and the time of perfusion, either by conventional intermittent dosing or extended or continuous infusion, may influence its PK.

β-lactam antibiotics act through a pathway that blocks the growth and replication of bacteria, leading to their death. Their main mechanism of action is the inactivation of enzymes located in the bacterial cell membrane, known as penicillin-binding proteins, which are involved in cell wall synthesis [33]. It is well known that different β-lactam antibiotics may preferentially bind to and inhibit specific penicillin-binding proteins according to their spatial structures [34], which leads to different efficiencies in the killing of microorganism. These antibiotics have bactericidal properties against replicating bacteria, meaning they kill the microorganisms through the inhibition of bacterial cell wall synthesis, which is necessary for replication, leading to their lysis after division is complete [33].

Although all β-lactams share the same β-lactam ring, they present different structures and bacteria-sensitive profiles, leading to their classification in different classes, namely penicillins, cephalosporins, monobactams, and carbapenems [35,36]. These changes in antibiotic structure are directly related to their stability against bacterial resistance mechanisms (especially β-lactamase enzymes).

These antibiotics are time-dependent, that is, their activity is mainly related to the time their concentration is above the minimal inhibitory concentration (MIC). Different thresholds have been determined for the diverse classes of β-lactams in vitro or in animal models [37]. Bacteria killing is ensured as soon as this threshold is attained but does not increase significantly with higher concentrations [38]. Moreover, no significant post-antibiotic effect is generally noted [38,39].

The concept of dose modulation consists of a broader and more innovative view of de-escalation. This includes the front-line variability in the antibiotic dosage, according to patient and microorganism characteristics, followed by its reduction after clinical response and patient recovery. This means concentrating the largest weight of antibiotics at the front end [40], when the microbial load is higher and the PK changes pose the highest risk of underdosing, and reducing the antibiotic dose when the sepsis syndrome is improving [22], guided by PK and PD data.

In fact, in patients responding to therapy and recovering from infection, there is commonly a hemodynamic improvement, with the weaning of vasopressors, a negative fluid balance, and the normalization of cardiac, renal, and hepatic functions, which also lead to progressive antibiotic PK normalization [28].

Therefore, after hemodynamic recovery, if the patient is improving and there is evidence of a clinical response, it is probably safe to reduce the antibiotic dose to reduce the risk of antibiotic toxicity. Furthermore, there is a large amount of evidence that the duration of antibiotic treatment can be largely decreased [41,42], which may help to control the overall antibiotic exposure and the emergence of resistance.

This strategy, along with the use of a broad-spectrum combination antibiotic therapy, may help to overcome the limitations to its efficacy posed by the large inoculum [43], the PK changes, and the decreased antibiotic susceptibility [44].

Conventional antibiotic dosing, accounting only for changes in renal or hepatic clearance, may be prone to failure in complex critically ill patients [24].

In fact, despite the PK changes, the dose is usually maintained throughout the treatment course. With an expanding knowledge of the relationships between antimicrobial dosing, pharmacokinetic/pharmacodynamic (PK/PD) exposure, and patient outcomes [22], there is now a strong rationale to individualize antimicrobial dosing in critically ill patients with the aid of TDM [10].

As already pointed out, antibiotic dosing in critically ill patients is especially challenging due to the increased Vd [21] and the rapid changes in renal and hepatic clearance. Moreover, bacterial inoculum, commonly much higher in critically ill patients, may also influence the bacteria-killing kinetics and may contribute to therapeutic failure [28]. Consequently, dose individualization according to the PK of the antimicrobial and the patient’s unique clinical characteristics may have a strong impact on achieving the optimal therapeutic exposure based on the bacterial susceptibility. Theoretically, this may help to maximize the killing of bacteria, minimize toxicity, and prevent the emergence of resistance, which is especially important when dealing with patients at a high risk of death and in environments with escalating antibiotic resistance [40].

Therefore, TDM for β-lactams has been suggested for routine implementation in the treatment of critical illness. Accordingly, TDM would help to implement the dose modulation strategy, facilitating the early attainment of the maximal tolerable dose [45] of the selected antibiotic. This would facilitate avoiding initial antibiotic underdosing on one hand, potentially improving the antibiotic effectiveness and, on the other hand, promoting biosafety, decreasing the antibiotic exposure time, and avoiding overexposure and toxicity [40].

The most conventional and practical method to use TDM is based on the evaluation of a single PK sampling at the end of a dosing interval (a trough sample with a minimal concentration). The measured serum concentration is compared against a therapeutic target range, and dosage adjustments are performed. This constitutes the easiest methodology, but it is also the least accurate for dosing adjustments. Limited sampling strategies incorporate up to three samples, providing more informative concentration time points for PK description [46,47]. These can be conjugated with dosing nomograms, which are ideally tested on similar populations [48], and can incorporate data on organ function, along with the PK/PD data, providing a more accurate method for dosing adjustment [21]. The dosing of the total concentration of the drug may be misleading and the measurement of the free drug concentration should be preferred [26].

Given the safety profile of β-lactams, TDM has been proposed to focus on maximizing efficacy through the achievement of adequate therapeutic exposures. As observational data suggested larger-than-conventional PK/PD targets, several TDM studies focused on achieving 100% values for the time with a free drug concentration above the MIC (100% fT > MIC) or even more than four times above that target [12,49]. The measurement of bacterial MIC values, drug sampling, and the use of nomograms may help to achieve these predefined therapeutic targets [29,50].

In a large study, assessing the TDM of β-lactams in 330 critically ill patients, Wong et al. were able to achieve these more ambitious therapeutic targets in 63.4% of patients [51]. However, in the population with the microbiological documentation of an infection, there were no differences in outcomes (odds ratio: 0.88 (95% CI: 0.40–1.91); p = 0.74 and odds ratio: 0.67 (95% CI: 0.29–1.55); p = 0.35 for 100% fT > MIC and 100% fT > 4 × MIC, respectively) [51].

The potential benefits of TDM are mainly related to the existence of a recognized relationship between a serum drug concentration and the intended effect. Moreover, this clinical benefit is mostly based on the possibility of having a rapid and inexpensive method to measure drug concentrations reliably and with a short turnaround time (allowing real-time dosing adjustment). As of today, this may not be the case for β-lactam antibiotics.

Dynamic changes in therapeutic targets according to the mode of antibiotic administration (either continuous infusion or conventional intermittent dosing) have been unveiled in an in vitro model [52]; furthermore, excessive reliance on MIC determination [53] may lead to false assumptions, which may explain the incongruencies between achieving PK/PD targets but attaining inadequate clinical outcomes.

There is increasing evidence that front-line antibiotic inappropriateness is common and may have a significant impact on the outcomes of patients with severe infections and septic shock [54]. An appropriate spectrum of antibiotic therapy may be insufficient if adequate exposure is missed [55]. However, the PK changes that occur in infected critically ill patients often lead to significantly different antibiotic exposures, which have an impact on their effectiveness.

In clinical practice, the major benefit of antibiotics, rapid bacteria killing, seems to be mostly concentrated to the first days of therapy. For β-lactams, the early achievement of 100% fT > MIC with continuous or extended infusions [56] during the first hours of therapy is probably safe and may improve clinical outcomes. Increasing the time of β-lactam perfusion is associated with extended fT > MIC [57], although the total exposure remains the same.

In critically ill patients, higher targets have been proposed, especially for β-lactams, including 100% of the time over a concentration as high as 4 × MIC [58]. Consequently, extended or even continuous infusions of these antibiotics have been proposed to be able to achieve these higher targets and improve clinical outcomes. However, theoretically, the plateau concentration of β-lactams, which is associated with continuous infusion, may be below the efficacy threshold 100% of the time, and this could lead to the worst outcome.

The strategy of the extended infusion of β-lactams was used in a study addressing piperacillin/tazobactam in patients infected with Pseudomonas aeruginosa. In this before–after study, a significant decrease in mortality in the most severe group of patients was noted (12.2% vs. 31.6%, respectively; p = 0.04) [59]. However, this very outcome was not found in other studies [60,61,62] or large meta-analyses [63,64], reinforcing the complexity of the PK/PD targets.

A meta-analysis of the individual data of three prospective studies comparing continuous infusion with intermittent doses of β-lactams [65] unveiled a small but significant benefit of continuous infusion in reducing in-hospital mortality (19.6% vs. 26.3%, relative risk: 0.74 (95% confidence interval: 0.56–1.00), p = 0.045). Some answers may come from a large multicenter trial with a similar protocol that is just finishing recruitment [66].

It is important to note that the limited data relating to PK/PD targets and outcomes should not be viewed as proof of a lack of benefit. While the attainment of any antibiotic concentration target does not guarantee an improved outcome per se, using individually guided dosing optimizes the probability of achieving adequate bacterial killing [67]. Variability in targets (such as the organisms’ MIC values), intra- and interindividual variability in PK, β-lactam “silent” toxicity, and possible differences in the adequate PK/PD targets according to the bacteria and type of infection point to the complexity of this process.

Furthermore, the measurement of the clinical response is also not standardized. Decreases in biomarkers [68,69] or even in the bacterial inoculum [70] may be used, but their impacts on antibiotic management are still unclear.