Pharmacokinetics–pharmacodynamics issues relevant for the clinical use of beta-lactam antibiotics in critically ill patients

Antimicrobials are among the most important and commonly prescribed drugs in the management of critically ill patients and beta-lactams are the most common antibiotic class used. Critically ill patient’s pathophysiological factors lead to altered pharmacokinetics and pharmacodynamics of beta-lactams. A comprehensive bibliographic search in PubMed database of all English language articles published from January 2000 to December 2017 was performed, allowing the selection of articles addressing the pharmacokinetics or pharmacodynamics of beta-lactam antibiotics in critically ill patients. In critically ill patients, several factors may increase volume of distribution and enhance renal clearance, inducing high intra- and inter-patient variability in beta-lactam concentration and promoting the risk of antibiotic underdosing. The duration of infusion of beta-lactams has been shown to influence the fT > minimal inhibitory concentration and an improved beta-lactam pharmacodynamics profile may be obtained by longer exposure with more frequent dosing, extended infusions, or continuous infusions. The use of extracorporeal support techniques in the critically ill may further contribute to this problem and we recommend not reducing standard antibiotic dosage since no drug accumulation was found in the available literature and to maintain continuous or prolonged infusion, especially for the treatment of infections caused by multidrug-resistant bacteria. Prediction of outcome based on concentrations in plasma results in overestimation of antimicrobial activity at the site of infection, namely in cerebrospinal fluid and the lung. Therefore, although no studies have assessed clinical outcome, we recommend using higher than standard dosing, preferably with continuous or prolonged infusions, especially when treating less susceptible bacterial strains at these sites, as the pharmacodynamics profile may improve with no apparent increase in toxicity. A therapeutic drug monitoring-guided approach could be particularly useful in critically ill patients in whom achieving target concentrations is more difficult, such as obese patients, immunocompromised patients, those infected by highly resistant bacterial strains, patients with augmented renal clearance, and those undergoing extracorporeal support techniques.


Background
Antimicrobials are among the most important and commonly prescribed drugs in the management of critically ill patients and beta-lactams are the most common antibiotic class used because of their broad spectrum of activity and high tolerability [1,2].
Early and appropriate antibiotic administration improves clinical outcome of septic patients [3][4][5][6][7]. In the presence of septic shock, besides conflicting results [8,9], each hour delay is associated with a measurable increase in mortality and other negative endpoints (e.g., length of stay in ICU, acute kidney injury, acute lung injury, and global organ injury assessed by the Sepsis-Related Organ Assessment score) [10,11].
Choosing the appropriate antimicrobial for the bacterial activity spectrum is crucial but the correct dosage regimen (both dose and frequency) is, at least, of the same importance for successful clinical cure and microbiological eradication [11].
Unlike organotropic drugs, where it is easy to titrate dose to achieve a clinical response, antibiotics may take 24-72 h to present signs of resolution of infection, making it difficult to determine the most appropriate dosage [1,2].
We conducted a comprehensive bibliographic search in the PubMed database of all English language articles published from January 2000 to December 2017, using the following keywords: critical care or intensive care or critically ill and sepsis or septic shock and antibiotics and pharmacokinetics or pharmacodynamics. Articles not addressing beta-lactam pharmacokinetics (PK) or pharmacodynamics (PD) in critically ill patients were excluded. A small number of articles derived from references in the articles selected were also reviewed. In the end, 214 studies were included in our review (Fig. 1).

Beta-lactam PD characteristics
Knowledge of the antimicrobial PD characteristics (inhibition of growth, rate and extent of bactericidal action, and post-antibiotic effect (PAE)) provides a more rational basis for determination of optimal dosing regimens in terms of the dose and the dosing interval.
The antimicrobial activity of drugs is usually assessed by determination of the minimum inhibitory concentration (MIC) and the minimum bactericidal concentration (MBC) of the drug under specific conditions in vitro. These in vitro conditions are very different from those expected at the site of infection, where the milieu is frequently acidic and anaerobic, and tissue protein may bind a variable amount of the drug. Additionally, these parameters do not provide information on the time course of the antimicrobial effect-the fluctuating levels that are present in a patient treated with the drug-and are measured against a standard bacterial inoculum (about 10 colony-forming units (CFU) per millilitre [5]) that does not necessarily correspond to bacterial densities at site of infection (10 CFU per gram of tissue or pus [8][9][10]). For bactericidal drugs, the MBC is generally not more than fourfold higher than the MIC. The size of the residual bacterial population at the end of each dosing interval, and ultimately the efficacy of the antimicrobial regimen, depends on the interplay of a variety of bacterial, drug, and host factors that include the size of the initial bacterial population, the potency (MIC and MBC) and PK characteristics of the antimicrobial agent, the rate and extent of any bactericidal effect, the presence of a post-antibiotic effect (PAE), the rate of re-growth of persistent organisms, and the state of host defences [13].
Three PD indices describe optimal killing associated with antibiotics: fT > MIC, which is the amount of time that the unbound drug concentration remains above the MIC of the infecting organism; C max /MIC, which is the ratio between the maximum concentration of the drug and the MIC of the infecting organism; and AUC 0-24 / MIC, which is the ratio between total area under the concentration-time curve (AUC) over 24 h and the MIC of the infecting organism.
Increasing drug concentrations much above the MBC does not enhance bacterial killing and the bactericidal action of these drugs is relatively slow. When drug levels at the site of infection fall below the MIC, the relatively large residual population can resume growth quickly because most beta-lactams either have no or only a short PAE [12]. McKinnon et al. compared the PD of cefepime and ceftazidime and observed that patients with fT > MIC of 100% had significantly greater rates of clinical cure and bacteriological eradication than patients with fT > MIC of < 100% [15].
It is suggested that 50% fT > MIC of the dosage interval is needed to ensure standard efficacy with these antimicrobials, whereas 100% fT > MIC of the dosage interval should be ensured for optimal exposure in immunocompromised patients. A further improvement in efficacy is observed when antibiotic concentrations are four to five times greater than MIC [2,12,13]. The percentage of time above MIC that correlates with efficacy varies among different beta-lactam groups, being greater for cephalosporins and aztreonam than for penicillins, and greater for penicillins than for carbapenems. Also, variations occur among different bacterial species, being less for staphylococci, for which beta-lactams have a PAE, than for streptococci and Gram-negative bacilli, for which beta-lactams do not have a PAE [2,13].
Beta-lactam PK issues in the critically ill Discovered antibiotics are evaluated in vitro and tested in animals, initially for toxicity and subsequently for efficacy. The antibiotic dose and frequency are based on these in vitro or animal in vivo PK/PD studies. These dosing regimens are then tested on healthy human volunteers for tolerability, with clinical efficacy studies undertaken in non-critically ill patients. After the launch of the drug onto the general market, the same dosing regimen is used in critically ill patients; however, this is likely to lead to suboptimal outcomes in the ICU [5], especially with more resistant bacterial strains [16] and in the immunocompromised population [17].
Beta-lactams are hydrophilic drugs and so their volume of distribution (Vd) is low and similar to that of extracellular water. Variations in the extracellular fluid content and/or in renal or liver function may be considered the most relevant and frequent pathophysiological mechanisms possibly affecting drug disposition in critically ill patients. Other factors may contribute to altered antibiotic concentrations: an interesting case-report by Taccone et al. [18] related the case of an obese septic patient with Pseudomonas aeruginosa pneumonia treated with meropenem. The PD target (t > 4 × MIC > 40% of the dosing interval) was only achievable by dosing 3 g q6h at 3 h extended infusion and was associated with clinical improvement.
PK studies on ICU septic patients reported, overall, increased Vd with significant daily concentration fluctuations between and within patients [5,[36][37][38][39][40][41]. Clearance of drugs is also affected and usually related to creatinine clearance [1,[42][43][44]. A single-center study of 17 ICU patients with ventilator-associated pneumonia (VAP) described the PK profile of ertapenem and concluded that, because of its highly protein-bound profile, hypoalbuminemia resulted in a higher protein-unbound fraction with consequences for drug distribution and elimination [38]. Ulldemolins et al. [39] found the same while studying the PK profile of flucloxacillin. Ramon-Lopez et al. [45] described high PK variations (between and within patients) for meropenem in 12 burn ICU patients that were mostly related to age, body weight, and serum albumin. Carlier et al. [37] investigated the adequacy of piperacilin/tazobactam dosing and its trough variability during an entire 7-day antibiotic course in 11 ICU patients with pneumonia and normal renal function. Six of them failed to achieve the PK/PD target of 100% fT > MIC at least once during the treatment course and considerable antibiotic concentration variability was found within and between patients. The DALI study, a large multicenter prospective study evaluated 248 ICU patients treated for infection with beta-lactams and found large variations on beta-lactam blood concentrations. The achievement of the PK/PD targets was highly inconsistent, with one fifth of the patients not achieving their most conservative PK/PD target of 50% fT > MIC and better outcomes were described with higher drug exposure, at least for less severely ill patients [5].
Septic patients with acute renal failure may have suboptimal antibiotic concentrations in the first days of therapy when the recommended dosing adjustment for renal failure is used [46]. Taconne et al. [40] studied the PK profiles of four beta-lactams (ceftazidime, cefepime, piperacilin/tazobactam, and meropenem) over the first 24 h of treatment in 80 septic ICU patients. They concluded that, besides high intra-and inter-patient PK variability, standard first doses of broad-spectrum β-lactams provided inadequate levels to achieve target serum concentrations for extended periods of time.
Augmented renal clearance has probably more impact than altered Vd on the PK of beta-lactams [25,27,[47][48][49][50][51][52][53][54]. Roberts et al. [23] described the PK of cefazolin in plasma and interstitial fluid of subcutaneous tissue in post-trauma critically ill patients and demonstrated that increasing creatinine clearance (ClCr) or decreasing serum albumin concentrations will reduce the likelihood of achieving optimal cefazolin exposure in subcutaneous interstitial tissue. In the presence of augmented renal clearance (ClCr > 130 mL/min), a much higher dose of cefazolin is required to obtain similar relative drug exposures [23]. Conil et al. [43] found that higher ClCr values (> 50 mL/min) did not provide trough concentrations of piperacilin (4 g three times a day) sufficient enough to attain the MIC for many pathogens in many of the patients studied.
Hypoalbuminemia has also been associated with altered PK. Wong et al. [55] described a linear correlation between the percentage protein binding of flucloxacillin and the plasma albumin concentration, though this was not true for ceftriaxone. Also, plasma albumin concentrations and in vitro binding data from healthy volunteers should not be used to predict unbound concentrations of ceftriaxone in ICU patients [56].

Use of extracorporeal support techniques in critical care
Acute kidney injury (AKI) occurs in 50 to 65% of critically ill patients and in approximately two-thirds of patients within the first 24 h after admission to the intensive care unit (ICU) [57]. Critically ill patients are usually supported with one of the forms of continuous renal replacement therapy (CRRT)-continuous venous-venous hemofiltration, hemodiafiltration, hemodialysis (CVVHF, CVVHDF, CVVHD, respectively)-or with sustained low-efficiency dialysis (SLED). Molecules are transported across the filter membrane by the mechanism of convection (driven by the pressure gradient-CVVHF), diffusion (driven by the concentration gradient-CVVHD, SLED), or both (CVVHDF).
Employing CRRT complicates antibiotic dosing to a significantly higher extent than standard hemodialysis due to the high number of variables, including Vd, flow of the dialysis fluid, replacement fluid infusion site (preor post-dilution mode), type and surface of the used membrane, and the difference between delivered and prescribed RRT dose.
Vd in AKI may be significantly different from published population estimates derived from healthy subjects. Besides the decreased plasma protein concentrations in acutely ill patients, uremic solutes, such as hippurate and indoxyl sulfate, alter drug binding to albumin in chronic renal failure and might do so in acute renal failure, although this has not been tested. The free fraction of many drugs is increased in renal failure, even though the Vd for total drug may increase due to movement of unbound drug into interstitial or total body water [57][58][59].
Overall, a tendency for antibiotic underdosing in critically ill patients on CRRT or SLED likely exists. The mode and dose of CRRT vary quite widely from center to center and from report to report, making it very difficult to create generally applicable beta-lactam dosing guidelines for critically ill patients under CRRT. Additionally, antibiotic concentrations may vary depending on the degree of extraction and residual renal function, which is variable, difficult to assess, and rarely considered despite its relevant contribution to antibiotic clearance in patients undergoing CRRT (Tables 1 and 2) .
Globally, we recommend not reducing standard antibiotic dosage since no drug accumulation was found in the available literature and to maintain continuous or prolonged infusion in critically ill patients on CRRT, SLED, or EDD, especially for the treatment of multidrug-resistant bacteria. Although usually not available in clinical routines, a therapeutic drug monitoring (TDM)-guided strategy has potential benefit to ensure appropriate antibiotic therapeutic targets.
Extracorporeal membrane oxygenation (ECMO) has become an essential tool for severe cardiorespiratory failure in critically ill patients. It is thought to introduce additional confounding factors to the already altered PK properties of beta-lactams in this subset of patients. Sequestration of antibiotics in the ECMO circuit and the associated systemic inflammation can further increase the antibiotic Vd and reduce clearance [74,[97][98][99]. However, very few in vivo studies have been performed in this subset of patients (Table 3). Globally, they show no significant statistical variation in Vd and clearance, but while probability of target attainment (PTA) with standard ICU dosage regimens was achieved when treating for highly susceptible Gram-negative bacteria, antibiotic concentrations were below those desired to treat more resistant strains.
Longer exposure regimens: continuous infusion, extended infusion, or reduced-interval dosing The duration of infusion of beta-lactams has been shown to influence their fT > MIC. Improved PD profiles of beta-lactams may be obtained by longer exposure with more frequent dosing, extended infusions, or continuous infusions. Several studies reported PD benefits for target attainment of extended and continuous infusions, especially considering highly resistant bacterial strains, even using smaller daily doses [1,2,36,41,.      However results are conflicting concerning decreased mortality and bacteriological and clinical cure rates [144]. A sub-analysis from the DALI study compared intermittent bolus vs prolonged infusions of beta-lactams in patients with respiratory infection and concluded that patients receiving beta-lactams via prolonged infusion demonstrated significantly better 30-day survival [145]. Falagas et al. [114] conducted a meta-analysis of 14 studies comparing continuous and short-term infusion of carbapenems and piperacilin-tazobactam, involving 1229 patients. Mortality was lower among patients receiving extended or continuous infusion of carbapenems or piperacillin/tazobactam compared to those receiving short-term infusion (risk ratio (RR) 0.59, 95% confidence interval (CI) 0.41-0.83). Patients with pneumonia who received extended or continuous infusion had lower mortality than those receiving short-term infusion (RR 0.50, 95% CI 0.26-0.96) [114].
An interesting retrospective study by Huang et al. [120] reviewed 68 neurosurgical patients with post-operative intracranial infections treated with 4 g/day cefepime over 24 h as a continuous infusion (CI; n = 34) or 2 g every 12 h as intermittent infusion (II; n = 34). CI controlled the intracranial infection more rapidly and effectively than II (6.6 ± 1.9 days vs 7.8 ± 2.6 days; P = 0.036). PD targets were more achievable with CI: for plasma cefepime concentrations, the percentage fT > MIC in the CI group was higher than in the II group (for MICs of 8 μg/mL, 100% vs 75%, respectively). For cerebral spinal fluid (CSF) cefepime concentrations, the percentage fT > MIC in the CI group was higher than in the II group (for MICs of 4 μg/mL and 8 μg/mL, 83.3% and 75% vs 25% and 0%, respectively) [120].
De Waele et al. [27] reviewed 343 patients from 68 ICUs across ten countries and concluded that use of intermittent infusion was the most significant factor associated with target non-attainment, for both 50% and 100% fT > MIC. Other risk factors for target non-attainment were ClCr, recent surgery, and timing from initial antibiotic therapy and sampling. However, the type of infusion was such a significant covariate in the model that it eliminated the effects of other variables [27].

Site of infection
Usually drug concentrations in blood are used to determine PD parameters, such as percentage of time drug levels exceed the MIC and peak drug AUC/MIC level, due to the relative accessibility of this body fluid. Because infection usually occurs at extravascular sites, the use of drug concentrations in blood is only satisfactory if blood levels are an adequate surrogate for levels at the site of infection [13]. In septic shock, blood misdistribution in the microcirculation might decrease antibiotic concentration at the infection site [1].
Boyadjiev et al. [146] studied ertapenem penetration into muscle in mechanically ventilated patients and concluded that average muscle free-ertapenem concentrations were above the MIC values of targeted pathogens except in a few patients. Karjagin et al. [147] evaluated the PK/PD relations of meropenem in plasma and peritoneal fluid by microdialysis and showed that area under the concentration-time curve was lower in peritoneal fluid than in plasma, concluding that in patients with severe peritonitis associated with septic shock, a dosing regimen of 1 g infused over 20 min every 8 h is sufficient against susceptible bacteria, but not always against intermediately susceptible bacteria. Also, beta-lactam PK is variable between plasma and subcutaneous interstitial fluid in septic patients [148]. Thus, prediction of microbiological outcome based on concentrations in plasma results in overestimation of antimicrobial activity at the site of infection.
Special anatomic barriers (e.g., brain, eye, and prostate) can result in drug levels being much lower than free drug levels in plasma [13]. The combination of tight junctions and active transport systems that form the blood-brain barrier creates a substantial impediment to the penetration of most antibiotics into the CSF. However, the presence of inflammation within the meninges significantly alters the permeability of the blood-brain barrier, increasing CSF exposure for the majority of antibiotics [20]. For meningitis, CSF levels are appropriate for determination of PD parameters.
Very few studies have investigated PK/PD issues in the CSF (Table 4). Five case reports, one randomized clinical trial in a paediatric population, and three prospective observational studies found good probability of target attainment for susceptible strains but standard dosing may not be optimal for less susceptible strains. Prolonged and/or continuous infusion is of benefit in the attempt to achieve PD targets. No data regarding intermittent versus continuous CSF ventricular drainage were found and conceptually these two types of drainage may alter the beta-lactam PK profile.
There is very sparse data on possible surrogate central nervous system penetration factors for beta-lactams, so no conclusions can be made. We recommend to use higher than standard dosing, preferably with continuous or prolonged infusions, especially when treating less susceptible bacterial strains. Toxicity did not increase at increased doses. Finally, none of these studies addressed clinical outcome.
Though there are PK models of plasma concentrations of beta-lactams specifically for the critically ill population with pneumonia, it is suggested that epithelial lining fluid (ELF) concentrations are important determinants of efficacy of treatment of bacterial pneumonia. ELF-to-serum penetration ratios may vary widely among   90 64 mg/L) It is very difficult to achieve this PK/PD standard in the CSF, and a higher dose might be needed to treat intracranial infections TBI trauma brain injury beta-lactams [13,20,159]. The impact of infection on their penetration into ELF in humans is unknown [159], though some reports state that ELF penetration increases in acute lung injury [160].
Only a few studies have investigated beta-lactam PK/PD issues in critically ill patients with pneumonia (Table 5) and in only seven of them were ELF drug concentrations measured. A standard dosage of beta-lactams derived from healthy patients' PK profiles may be insufficient for treatment of critically ill patients with pneumonia, especially when caused by multidrug-resistant pathogens. Continuous or prolonged infusions and higher than standard doses improve the PD profiles of these antibiotics. This is very important to achieve an adequate PD profile when treating less susceptible bacterial strains. Therapy drug monitoring would be extremely helpful in this setting.

New beta-lactam drugs and beta-lactamase combinations
Of great concern is the worldwide increase in the number of infections caused by Gram-negative multidrug-resistant bacteria. Treatment choices for these infections have been limited, especially for infections caused by bacteria that produce carbapenemases and/or extended-spectrum beta-lactamases.
Ceftolozane-tazobactam and ceftazidime-avibactam are 2 beta-lactams/beta-lactamase combinations with anti-Gram-negative bacteria activity that were recently approved for the treatment of complicated intra-abdominal infections, complicated urinary tract infections, and nosocomial pneumonia.
Ceftolozane is an oxyimino-aminothiazolyl cephalosporin with a pyrazole substituent at the 3-position side chain instead of the lighter pyridium present in ceftazidime. This heavier side chain provides improved steric hindrance to prevent hydrolysis mediated through AmpC beta-lactamases.
Ceftolozane-tazobactam combines a novel cephalosporin with an established beta-lactam beta-lactamase inhibitor, whereas ceftazidime-avibactam couples a well-known cephalosporin with a novel non-beta-lactam beta-lactamase inhibitor.
Both tazobactam and avibactam target the active site of serine beta-lactamases. Tazobactam, a beta-lactam sulfone, binds irreversibly to the active site of beta-lactamases and avibactam is a diazabicyclooctane non-beta-lactam that binds covalently and reversibly to beta-lactamases. This reversibility is a unique feature that allows avibactam to undergo recyclization to inactivate another beta-lactamase. The crucial advantage of avibactam is its ability to inhibit extended spectrum beta lactamases, AmpC beta-lactamases (as expressed in Pseudomonas aeruginosa and Enterobacteriaceae), and class A carbapenemases of the Klebsiella pneumoniae carbapenemase (KPC and OXA-48) family.
The pharmacokinetic and safety profiles of this antibiotic have been established in healthy adults and subjects with various degrees of renal function [170,171]. The currently approved dosages for adult patients with an estimated ClCr > 50 mL/minute are ceftolozane 1 g with tazobactam 500 mg every 8 h and ceftazidime 2 g with avibactam 500 mg every 8 h for complicated urinary tract infections and intra-abdominal infections [172] and ceftolozane 2 g with tazobactam 1 g every 8 h for nosocomial pneumonia [173].
However, data guiding its use in critically ill patients are currently sparse, being entirely derived from studies with very few patients and/or case reports.
Veillete et al. [174] presented PK data for ceftazidimeavibactam from two patients with bloodstream infections caused by carbapenemase (KPC)-producing K. pneumoniae; the patients had renal impairment and one of them was obese. In both patients half-lives were prolonged and Vd larger than predicted. They conclude that recommended doses and intervals may not be sufficient for obese patients with renal failure, especially for those infected with KPC-producing organisms [174].
Oliver et al. [175] evaluated the adequacy of extended-infusion ceftolozane-tazobactam to achieve target PK and PD goals in a critically ill patient with Pseudomonas aeruginosa pneumonia and septic shock on CVVH. A dosage of 1.5 g every 8 h (3-h infusion) was given. All estimated plasma-free drug concentrations achieved the PD goals and remained well above the isolated organism's MIC of 1.5 μg/mL and above the susceptibility breakpoint of 4 μg/mL throughout the dosing interval, although the authors could not comment on drug concentrations at the site of infection. The authors conclude that, given the lowest estimated free-drug concentration was fivefold greater than the susceptibility breakpoint, the estimated half-life of 28 h and the low extraction ratio observed, a lower total daily dose might be utilized and an extended infusion time may not be necessary for patients on CVVH [175].
Bremmer et al. [176] performed a PK analysis of intravenous ceftolozane-tazobactam 3 g every 8 h in a critically ill patient with P. aeruginosa pneumonia on CVVHDF. They concluded that, compared with a patient with normal renal function, this patient had decreased ceftolozane clearance. A ceftolozane-tazobactam dosage of 1.5 g every 8 h should adequately achieve a desired drug concentration above the minimum inhibitory concentration of 8 μg/mL for the treatment of pneumonia [176].
Stokem et al. reported the successful treatment with ceftolozane-tazobactam 3 g every 12 h for a pulmonary exacerbation in a 35-year-old female post-lung transplant, with cystic fibrosis, malnutrition, chronic kidney disease, and multi-drug resistant P. aeruginosa infection. Optimal time above MIC (estimated 100% time above    MIC of ceftolozane achieved against both isolates was 2 and 0.5 μg/mL) was likely attained at the dose and frequency provided in this case [177].

Toxicity
Beta-lactams are generally considered to have a high safety window with relatively few adverse effects, even when high doses are used [15]. Neurotoxicity is the most reported serious adverse effect of beta-lactams. Benzylpenicillin, cefepime, ceftazidime, and imipenem are considered to be the high-risk beta-lactams for neurotoxicity. Renal impairment, excess doses and/or concentrations, age, and a prior history of neurological disorders are known to be predisposing factors [2,[178][179][180][181][182][183][184].

Therapeutic dose monitoring
Several studies reported high PK variability of beta-lactams in sepsis/septic shock, both in different patients and in the same patient over time. In critically ill patients, hydrophilic and moderately lipophilic antimicrobials, being at higher risk of daily PK variations, should be more closely monitored and their dosages should be streamlined according to the underlying diseases in order to prevent under-or overexposure [2,11].
Therapeutic drug monitoring (TDM) has been instituted for aminoglycosides and glycopeptides to reduce the rate of toxicity. However, because of the safety profile of beta-lactams, TDM was thought unnecessary for these drugs. In line with PK changes in critically ill patients, insufficient PD target attainment with beta-lactams has been reported in these patients, especially those with hypoalbuminemia, altered renal function, and low susceptibility bacterial strain infections [2,35,42,198]. The challenges in achieving 'optimal' drug concentrations in the critically ill suggest beta-lactam TDM as a useful strategy to optimize drug exposure [199].
The TDM approach could be particularly useful in a certain group of critically ill patients in whom achieving target concentrations is more difficult, such as those with highly resistant bacterial strains, obese patients, immunocompromised patients, those undergoing renal-replacement therapies, and patients with augmented renal clearance [2,198,200,201].
Though there are PK models to estimate antibiotic concentrations over a range of creatinine clearance (CrCl) and on renal replacement therapy [67,[202][203][204][205], the use of CrCl as a tool to optimize beta-lactam dosing may not be reliable; although CrCl was significantly correlated with concentrations and clearance of broad-spectrum beta-lactams, changes in CrCl and RRT parameters do not reliably predict variations in drug PK/PD. In this setting, routine TDM should be considered to adapt beta-lactam doses [206].
Daily TDM of beta-lactams with dose adaptation in critically ill patients improves PD target attainment [207,208]. Case reports have shown that TDM improved clinical outcome [209], but the clinical efficacy of using drug levels to achieve adequate concentrations had never been properly evaluated [1,35,210,211] and there are reports concerning cost-effectiveness [111].
Facing poor implementation in beta-lactam TDM, Delattre et al. [212] proposed a predictive PK performance between an aminoglycoside and a beta-lactam. Due to physicochemical and PK similarities between aminoglycosides and beta-lactams, optimization of the beta-lactam dosage could be reached without any beta-lactam measurements, using TDM-related data of an aminoglycoside. The study aimed to characterize the PK of four beta-lactams (piperacillin, ceftazidime, cefepime, and meropenem) at the first dose in 88 critically ill septic patients co-medicated with amikacin, and to confirm the predictive performance of amikacin data on these PK, on a larger patient cohort, using a nonlinear mixed-effects modeling approach. There was a significant relationship between the exposure to amikacin and to beta-lactams. The population model presented was able to guide dosage adjustments for piperacillin, ceftazidime, cefepime, and meropenem during the early phase of severe sepsis in critically ill patients, using renal biomarkers or TDM-related aminoglycoside data [212].

Conclusions
The duration of infusion of beta-lactams has been shown to influence their fT > MIC and an improved PD profile of beta-lactams may be obtained by longer exposure with more frequent dosing, extended infusions, or continuous infusions. This is particularly relevant in the critically ill patient, as Vd and ClCr are often increased, namely in the early phase of systemic hyperinflammatory states, promoting the risk of antibiotic underdosing.
The use of extracorporeal support techniques, either for renal replacement or ECMO, may further contribute to this problem and consequently concentrations below those expected are often found for beta-lactams. Given the heterogeneity of extracorporeal support therapy modes, it is difficult to suggest a specific dosage, but we recommend not to reduce dosage since no drug accumulation was found in the available literature and to use continuous or prolonged infusions to achieve the adequate PD profiles necessary to successfully treat infections caused by less susceptible strains.
More studies are needed to define optimal dosing of new beta-lactams and new beta-lactam/beta-lactamase combinations, which are increasingly important to effectively treat multidrug-resistant bacterial strains, namely in patients on extracorporeal support therapy and with difficult-to-treat sites of infection.
Although, it is not currently a clinical routine in most hospitals and its clinical efficacy has not yet been properly evaluated, a beta-lactam TDM approach with daily dose adaptation, allowing personalized antibiotic dosing, should be particularly useful in critically ill patients in whom achieving target concentrations is more difficult, such as obese patients, the immunocompromised, patients with augmented renal clearance, those undergoing extracorporeal support therapy, or those infected with highly resistant bacterial strains. Studies comparing TDM-versus non-TDM-based beta-lactam regimens should be promoted.
However, infection usually occurs at extravascular sites and prediction of outcome based on antibiotic plasma concentrations may result in overestimation of antimicrobial activity at the site of infection. Very few studies have investigated PK/PD issues concerning special anatomic barriers like the brain and lung, but most suggest that standard ICU dosing for beta-lactams may be insufficient for low susceptibility/high MIC pathogens in these sites. Therefore, although no studies have assessed clinical outcome, we recommend using higher than standard dosing, preferably with continuous or prolonged infusions, when treating severe infections caused by less susceptible bacterial strains at these sites, as PD profiles may improve and toxicity does not seems to increase. Availability of data and materials All data generated or analyzed during this study are included in this published article (see bibliography).
Authors' contributions RV and JAP made substantial contributions to conception and design, acquisition of data, and analysis and interpretation of data; were involved in drafting the manuscript or revising it critically for important intellectual content; gave final approval of the version to be published; agreed to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.