Introduction

Optimal antibiotic dosing in the ICU is still a controversial issue that clinicians face daily. Despite compelling evidence supporting early and appropriate antibiotic therapy as one of the most effective interventions for improving patient outcome [1], antibiotic selection and dosing is often challenging in critically ill patients because of disease complexity, resulting physiological alterations, and reduced antibiotic susceptibilities of nosocomial pathogens. Besides, selecting an antimicrobial to which the causal agent is susceptible is not sufficient to achieve the best clinical outcomes, and factors such as adequate tissue penetration and achievement of a pharmacodynamic target associated with therapeutic success according to the antibiotic class are crucial for improving infection cure and patient morbi-mortality [2–4].

Beta-lactam antibiotics are time-dependent antibiotics, meaning that they exhibit optimal killing activity when plasma concentrations are maintained above the minimum inhibitory concentration of the bacteria during a percentage of the dosing interval (%T>MIC). Beta-lactams are also the most prescribed antibiotics in the ICU [5]. Significant and unpredictable pharmacokinetic variability of this pharmacological group has been well documented in critically ill patients: the volume of distribution (Vd) and the clearance (CL) of beta-lactams have been found to vary significantly depending on patient severity, proteinemia and organ failure, among other factors [3, 6]. Acute kidney injury and the requirement of continuous renal replacement therapy (CRRT) add further variability to beta-lactam CL. However, available clinical evidence supporting beta-lactam dosing in critically ill patients with septic shock and CRRT is not yet optimal, since recommendations are mainly elucidated from healthy volunteers’ data and from clinical studies with important patient variability and limited sample sizes.

The aims of this article are to describe the current clinical scenario of beta-lactam dosing in critically ill patients with septic shock and CRRT, to highlight the sources of variability among the different studies that reduce extrapolation to clinical practice, and to identify the opportunities for future research and improvement in this field. For this purpose, two of the most frequently prescribed beta-lactams for nosocomial infections (meropenem and piperacillin) and a highly protein-bound antibiotic usually prescribed for community-acquired infections (ceftriaxone) were chosen for a thorough review. A systematic review of all available data on beta-lactam antibiotic pharmacokinetics in critically ill patients with CRRT was beyond the scope of this article, as this has been done elsewhere [7–9].

Search strategy and selection criteria

Data for this review were identified by systematic searches of PubMed (1966 to November 2013), as well as references cited by relevant articles. Search terms included were ‘meropenem’ or ‘piperacillin’ or ‘ceftriaxone’, ‘critically ill patient’ or ‘intensive care unit’ or ‘critical illness’, ‘continuous veno-venous hemodiafiltration’ or ‘continuous veno-venous hemodialysis’ or ‘continuous veno-venous hemofiltration’ or ‘continuous renal replacement therapy’, and ‘pharmacokinetics’ or ‘pharmacodynamics’. Relevant articles written in English, Spanish and Catalan where considered for this review. Those articles describing the pharmacokinetics of meropenem, piperacillin/tazobactam and ceftriaxone in adult critically ill patients receiving CRRT were included.

Effect of septic shock and CRRT in antibiotic dosing optimization

Classically, the in vitro susceptibility of the causal pathogen has been the cornerstone of antibiotic prescription. However, selection according to susceptibility is only a component of the optimal antibiotic therapy, and many other factors must also be considered. In terms of posology, it is paramount to design dosing strategies that maximize the likelihood of attaining the pharmacodynamic target associated with therapy success in the biophase. This is complex in the critically ill patient with septic shock and CRRT since it is well known that critical sickness and clinical interventions can drive to physiological changes likely to alter drug pharmacokinetics [3] and therefore likely to compromise the attainment of these pharmacodynamic targets.

There are two important time periods that must be considered for antibiotic dosing. The first period corresponds to the first day of therapy, where the main determinant for dosing must be the Vd since this determines the early attainment of antibiotic concentrations within the therapeutic range. In critically ill patients with sepsis, increased Vd must be expected for hydrophilic antibiotics such as beta-lactams (see Tables 1, 2, 3, 4, 5 and 6), aminoglycosides and glycopeptides [10–38]. This increase may be due to the presence of bacterial endotoxins in the bloodstream, which has a cascade effect on the production of endogenous molecules that act on the vascular endothelium, leading to vasodilation and transcapillary leakage of fluid and proteins into the extracellular space, where these antibiotics distribute. When the Vd is abnormally increased, distribution of hydrophilic antibiotics such as beta-lactams becomes more extensive for trying to compensate this larger space, with greater movement of the drug molecules from the central compartment (bloodstream) to the peripheral compartments (mainly extravascular fluid). The amount of the drug in plasma consequently decreases, and therefore the plasma concentration decreases. Consequently, given a particular minimum inhibitory concentration, shorter %T>MIC values can be expected, which in turn may compromise beta-lactams’ pharmacodynamic target attainment [39]. Critically ill patients may therefore require front-loaded doses of beta-lactam antibiotics during the first 24 to 48 hours, regardless of organ function, in order to compensate the increased Vd and to reach concentrations within the therapeutic range on the first day of therapy [39].

Table 1 Available data on meropenem pharmacokinetics in continuous renal replacement therapy
Full size table
Table 2 Available data on meropenem pharmacokinetics in continuous renal replacement therapy
Full size table
Table 3 Available data on piperacillin pharmacokinetics in continuous renal replacement therapy
Full size table
Table 4 Available data on piperacillin pharmacokinetics in continuous renal replacement therapy
Full size table
Table 5 Available data on ceftriaxone pharmacokinetics in hemofiltration
Full size table
Table 6 Available data on ceftriaxone pharmacokinetics in hemofiltration
Full size table

The particular case of CRRT requirement poses another scenario where loading doses may be considered. At the time of CRRT initiation, antibiotic concentrations over time are in steady-state equilibrium (if the antibiotic was initiated before CRRT commencement), but one can hypothesize that the change in drug CL induced by CRRT initiation may lead to the breakage of this equilibrium and, consequently, to a decrease in drug concentrations. A new steady state will follow after seven half-lives since the introduction of the foreign source of drug CL. During this time period, however, concentrations may fall below the therapeutic range. At this point, an additional loading dose may help in the maintenance of therapeutic levels. This phenomenon of steady-state breakage follows the theoretical pharmacokinetics principles but there are no studies yet that describe it in critically ill patients and hence concrete loading dose recommendations cannot be provided. Certainly this is a very interesting area that deserves further research to be properly understood.

The second period starts from day 2. During this period, the estimated drug CL is the main determinant of dosing, with the objective of maintaining the equilibrium between input and output as the tissues should already hold therapeutic antibiotic concentrations. In this context, CRRT represents a particular challenge in terms of dosing, especially for hydrophilic antibiotics, as concentrations may vary depending on the degree of extraction, which in turn depends on the CRRT modality, on drug physicochemistry and, presumably, on CRRT intensity [7]. Moreover, residual renal function is usually variable, difficult to assess and rarely considered when dosing, despite its relevant contribution to antibiotic CL in patients undergoing CRRT that has been described for meropenem and piperacillin among others [26, 29, 32]. Finally, the patient’s condition evolves throughout the ICU stay so the influence of the previously mentioned factors may vary over time, making it difficult to generalize recommendations only based on CRRT modality and intensity. Dosing should ideally be titrated daily depending on the CRRT settings and the evolution of the patient’s renal function. With this aim, therapeutic drug monitoring (TDM) of trough levels might be a useful tool for refining dosing decisions during the maintenance phase of therapy, as it is routinely performed with aminoglycosides and glycopeptides. However, despite emerging data suggesting that beta-lactam TDM might improve the attainment of pharmacodynamic targets associated with therapeutic success [40], the impact of systematic TDM on clinical outcomes and resource use is still to be prospectively validated. Due to the variable pharmacokinetics of these drugs in critically ill patients with CRRT, TDM certainly deserves further investigation.

Determinants of drug clearance by CRRT

Among the many options for renal replacement, CRRT is the most used in the critical care setting due to its advantages in hemodynamically unstable patients compared with intermittent techniques [41]. Drug clearance through CRRT is multifactorial and depends on both drug characteristics and CRRT modality and intensity. Continuous venovenous hemodialysis is based on the principle of diffusion of solutes across a semipermeable membrane driven by a concentration gradient, while continuous venovenous hemofiltration clearance is driven mainly by convection removal, where a positive hydrostatic pressure drives water and solutes across the filter membrane from the blood compartment to the filtrate compartment, from which it is drained. Continuous venovenous hemodiafiltration is the most efficient technique for solute removal, consisting of a combination between the two abovementioned techniques and resulting in the removal of hydrophilic solutes with simultaneous water elimination [7].

Regardless of the modality prescribed, a common determinant of drug clearance in CRRT is protein binding. Due to protein size and electrical charge, protein-bound molecules are unable to pass through the filter membranes and only unbound molecules will be available for elimination by CRRT. This is so critical that both sieving coefficients and saturation coefficients are usually simplified as the unbound drug fraction. However, antibiotic protein-binding alterations have been broadly observed in ICU patients [6] due to the altered plasmatic protein homeostasis associated with critical illness (the SAFE study reported that 40 to 50 % of the ICU patients had albumins <25 g/l) [42] and due to the presence of other highly protein-bound exogenous drugs and endogenous molecules (such as bilirubin) in plasma. This may consequently translate into alterations in the extent to which an antibiotic is cleared by CRRT. However, whereas the effect of hypoalbuminemia on antibiotic pharmacokinetics in critically ill patients with preserved renal function has been documented in previous studies [6], there are no available studies regarding its combined impact with CRRT.

Another factor likely to affect the extent to which drugs are cleared by CRRT is the CRRT intensity. The question of what is the optimal CRRT intensity has been a controversial issue since its first implantation. Several studies have evaluated the impact of using different CRRT intensities on mortality and recovery of renal function in critically ill patients, with different, usually debatable, results [43–48]. Due to this lack of definitive evidence, current clinical recommendations define the area of best practice for CRRT intensity as lying between 20 and 40 ml/kg/hour [41], the clinician being responsible for individualizing the appropriate CRRT intensity for each particular patient. However, the impact of different CRRT intensities on antibiotic dosing requirements has not yet been sufficiently evaluated.

Additional to the abovementioned points, more variability in drug CL by CRRT may be introduced by medical devices that may coexist with CRRT in patients with septic shock, such as polymyxin B fiber columns (to reduce endotoxin levels in sepsis) or extracorporeal membrane oxygenation. Other factors such as filter lifespan, filter anticoagulants such as citrate and drug recirculation may also have an effect on drug CL. However, their potential for antibiotic adsorption and removal has not yet been estimated.

Main limitations of available pharmacokinetic studies

To discuss the current scenario of beta-lactam dosing in patients with septic shock and CRRT, we performed a thorough review of the existing clinical data for three of the most frequently used (and studied) beta-lactam antibiotics in the ICU. Tables 1, 2, 3, 4, 5 and 6 summarize the available evidence on meropenem, piperacillin/tazobactam and ceftriaxone pharmacokinetics in critically ill patients with CRRT [15–38, 49, 50].

Critical review of these studies has lead to identification of the following points that limit applicability of dose recommendations to critically ill patients with septic shock and CRRT.

Patient population

The identified studies handle a highly heterogeneous patient population, which may jeopardize the generalizability of the results. For example, there are studies that pool together patients with septic shock and cardiogenic shock [17, 31]. The physiopathology of these two types of shock, however, is very different: septic shock is caused by peripheral vasodilation, systemic inflammation and, consequently, increased Vd; while cardiogenic shock involves peripheral vasoconstriction, which should have no effect on the Vd.

Other studies include septic and polytrauma patients requiring CRRT [25, 32]. Of note, one of these studies overcame the admission diagnosis-driven variability by developing a population pharmacokinetics model. The investigators found that admission diagnosis significantly influenced pharmacokinetic parameters: trauma patients exhibited higher Vd and CL than septic patients (d = 69.5 and 15.7 l in trauma patients and septic patients, respectively; CL = 54.15 and 8.04 l/hour in trauma patients and septic patients, respectively) [25]. Patients with sepsis/severe sepsis may also substantially differ from patients with septic shock: septic shock patients may exhibit higher Vd due to capillary leakage and aggressive fluid resuscitation as compared with critically ill patients without septic shock. In spite of this, some of the available studies include patients with sepsis/severe sepsis and acute kidney injury [21, 33–35, 37, 49, 50] but not those with septic shock.

Furthermore, a significant number of the articles do not report clinical severity scores for the studied population. In particular, increasing Acute Physiology and Chronic Health Evaluation II scores have been shown to correlate with increased Vd for hydrophilic antibiotics such as aminoglycosides [12]. However, variations in the Vd of meropenem and piperacillin have been reported in the literature (see Tables 1, 2, 3, 4 and 5). Similarly, CRRT may be prescribed in patients who still present a significant residual renal function. The influence of residual renal function on piperacillin pharmacokinetics in patients receiving continuous venovenous hemofiltration has been assessed by Arzuaga and colleagues, and significant differences in piperacillin CL have been reported; for example, total drug CL in patients with creatinine CL > 50 ml/minute was tripled as compared with patients with creatinine CL < 10 ml/minute [51]. These points suggest that the one-size-fits-all dosing recommendations based only on CRRT prescription may not apply to all different types of critically ill patients, as they are a highly heterogeneous population that may require different doses.

Continuous renal replacement therapy modality and flow rate

Regarding CRRT modalities, there is discordance in the literature on whether a specific modality makes a difference or not in terms of dosing. While some studies support a difference in CL partially due to CRRT modality [49, 50], some others suggest that there are no substantial variations between modalities [22]. Theoretically, convective and diffusive methods eliminate molecules from the bloodstream using different processes, and therefore the total drug CL should differ between CRRT modalities, as has been shown with piperacillin and meropenem [49, 50], but a significant volume of dosing recommendations are still generic for CRRT.

Regarding CRRT intensity, emerging evidence suggests that the total flow rate affects the CL of hydrophilic drugs with low protein binding. For example, Beumier and colleagues developed a population pharmacokinetics model for vancomycin administered as a continuous infusion in critically ill patients with sepsis and septic shock, and found that inclusion of CRRT intensity as a covariate on CL significantly improved the model [52]. Similarly, a study by Bilgrami and colleagues specifically targeted patients with high-intensity CRRT (>4 l/hour) receiving meropenem and found that total drug CL was higher compared with previous studies with lower intensity CRRT, intensity being the main parameter that accounted for the differences in meropenem CL (R2 = 0.89) [15]. The high CRRT intensity was such a determinant of meropenem CL that the doses required for the coverage of less susceptible bacteria (minimum inhibitory concentration = 4 mg/l) were similar to those used in patients without renal failure (1,000 mg every 8 hours). These data suggest that different CRRT intensities may translate into different drug CL and therefore into different dose requirements. Importantly, one must also highlight that most of the published studies use CRRT intensities in the lower range of the area of best practice (1 to 2 l/hour; 14.3 to 28.5 ml/kg/hour for a 70 kg adult) [16, 17, 19–21, 27, 29, 30, 32, 34, 35, 49, 50], while the actual tendency in the clinical setting may be using CRRT intensities in the higher range (>30 ml/kg/hour), especially for septic patients [41, 46]. In fact, a recent study by Varghese and colleagues studied the pharmacokinetics of piperacillin/tazobactam in critically ill patients with anuria/oliguria and CRRT at a median intensity of 38.5 ml/kg/hour, and reported higher drug CL (median 5.1 (interquartile range 4.2 to 6.2) l/hour) compared with other studies that used lower CRRT intensities (see Table 3 and 4) [38].

Moreover, the methodology for the calculation of CRRT intensity is not defined in most of the studies. Some of the studies report that an absolute CRRT intensity was prescribed to all patients, without being normalized to body weight. This leads to inherently variable CRRT doses, inversely proportional to the actual patient’s weight. For instance, an absolute CRRT intensity of 2 l/hour for a 100 kg patient results in a relative flow rate of 20 ml/kg/hour, whereas for a 50 kg patient the rate is 40 ml/kg/hour. When relative flow rate is prescribed, clinicians usually use body weight previous to admission or ideal body weight, and calculate the flow rate using the following formula:

Flow rate = Q D + Q R / weight kg

where QD is the dialysis fluid flow rate (ml/hour) and QR is the replacement fluid flow rate (ml/hour).

The rationale of this methodology is to avoid variations in the calculated flow rate over time as the patient real weight fluctuates during the ICU stay (for example, due to fluid therapy or edema) [53]. However, most of the studies do not report how body weight was considered in spite of the fact that it is essential to know which CRRT intensity was prescribed [43]. When real body weight is used, the calculated flow rate may be falsely low, as the denominator in the equation usually increases during the ICU stay. Recommendations include application of body weight previous to admission or ideal body weight [43]. However, considering the increasing prevalence of obesity in developed countries, one should discuss whether ideal body weight or body weight previous to admission should be used.

Pharmacodynamic target for dosing recommendations

Antibiotic dosing recommendations intend to achieve a pharmacodynamic target that, for beta-lactams, is defined by the %T>MIC value [54]. Classical studies report that penicillins and monobactams require at least a 50 to 60 % T>MIC for maximal bactericidal activity, cephalosporins require a 60 to 70 % T>MIC and carbapenems require a 40 % T>MIC[54]. However, most of these recommendations are based on in vitro studies and on animal models of bacteremia, where penetration into the site of infection is not considered. In vivo, higher %T>MIC values in plasma may be needed for achieving the abovementioned targets in biophases other than the bloodstream, since penetration into the target site follows diffusion kinetics and depends on the physicochemistry of each particular tissue. For instance, Roberts and colleagues reported that continuous infusion of full doses of meropenem (that is, 100 % T>MIC in plasma) was required for achieving 40 % T>MIC for less susceptible pathogens in subcutaneous tissue [11]. Also, the attainment of a particular percentage of T>MIC may be modified by the susceptibility cutoff values for the different bacteria, which vary depending on the country where the study is performed (for example, European Committee on Antimicrobial Susceptibility Testing vs Clinical and Laboratory Standards Institute breakpoints). The recommendations based upon a particular minimum inhibitory concentration in Europe may therefore not apply to the United States of America and vice versa.

Critical review of clinical pharmacokinetics data leads to the final consideration that there are multiple missed opportunities in the available literature. Further studies should be more focused on the study population of critically ill patients with septic shock in order to avoid variability derived from pathophysiological conditions other than septic shock. Inclusion and exclusion criteria should therefore carefully evaluate the admission diagnosis and the patient condition during the study period. Also, a population pharmacokinetics approach would be preferred to the noncompartmental approach, since the noncompartmental approach draws inaccurate conclusions because covariates that have an effect on parameter variability cannot be identified. Finally, consensus regarding clinical pharmacodynamic targets for beta-lactams would be helpful in the unification of dosing recommendations.

Conclusions

Optimization of beta-lactam therapy in CRRT is complex and is dependent on several drug, CRRT and patient-related factors. Consideration of drug physicochemistry and protein binding, CRRT settings and disease-related pharmacokinetic alterations is essential for individualizing dose regimens with the purpose of attaining pharmacodynamic targets associated with success.

During the first day, an initial loading dose is required to achieve drug concentrations within the therapeutic range early in time, regardless of impaired organ function. This principle may also apply to the moment of CRRT commencement, where a loading dose may be required to maintain concentrations within the therapeutic range. From day 2, dosing must be adjusted to CRRT settings and residual renal function. The complexity of dosing occurs due to the great variability encountered. As such, TDM of trough levels of beta-lactams may be regarded as a promising and key tool to individualize dosing daily and to ensure optimal exposure to the antibiotic.

Current dose recommendations are based on studies with some drawbacks that limit their applicability to the current clinical scenario. Mainly, dosing recommendations in CRRT follow a one-size-fits-all fashion, despite emerging clinical data suggesting that beta-lactam CL is partially dependent on CRRT modality and intensity. Moreover, heterogeneous populations have been pooled in the studies, limiting extrapolation to critically ill patients with septic shock and CRRT. Finally, there is still some controversy on the %T>MIC value that must be chosen as the pharmacodynamic target associated with success for tailoring dosing recommendations.

Further research on dose adjustment of beta-lactam antibiotics in critically ill patients with septic shock and CRRT is required in order to establish reliable and up-to-date recommendations that ensure optimal therapy and thus increase the likelihood of optimal outcomes in this population.