Introduction
Despite the application of guideline-concordant antimicrobial therapy, severe infections still account for high mortality rates among critically ill patients [
1,
2]. Insufficient antibiotic exposure or failure to attain the pharmacokinetic/pharmacodynamic (PK/PD) target has been associated with worse clinical outcomes [
3‐
5]. However, adequate antibiotic dosing in critically ill patients is extremely complex, owing to pathophysiological changes and reduced antibiotic susceptibility to the pathogen [
6,
7]. PK/PD target attainment of up to 60% has been reported for beta-lactam antibiotics in critically ill patients [
7,
8].
Flucloxacillin is widely used to treat infections caused by Gram-positive bacteria [
9]. In critically ill patients, flucloxacillin exhibits variable plasma protein binding, ranging from 28 to 97% [
10,
11]. Maximal in vivo bactericidal activity of flucloxacillin and suppression of antimicrobial resistance can be expected when unbound serum concentrations exceed four times the minimum inhibitory concentration (MIC) for 50 to 100% of the dosing interval (ƒT
>4×MIC = 50–100%) [
12‐
15]. These high concentrations are required to treat more resistant pathogens and facilitate penetration of sufficient unbound flucloxacillin to the infectious extravascular regions in critically ill patients [
2,
4,
13‐
18].
Previous studies assessing PK/PD target attainment of unbound flucloxacillin in critically ill patients were heterogeneous, considering study population selection and reported target attainment percentages [
10,
11,
19‐
21]. Two studies reported over 99.9% target attainment for daily doses up to 12 g [
11,
20], whereas others indicated only 26–91% target attainment [
10,
19,
21]. Previous study populations consisted of patients with serum hypoalbuminemia (≤ 32 g/L) [
10,
20], or reported median estimated glomerular filtration rates (eGFRs) of at least 96–122 mL/min [
10,
20,
21]. In addition, most previous studies have reported study population ages of up to only 59 years [
10,
11,
20,
21], considered non-representative of critically ill patients [
2,
22]. Moreover, the median non-coronavirus disease (COVID-19) age was 67 years in Dutch critically ill patients [
2].
Considering the above-listed findings, we performed a population PK multicenter study in a study population with widely ranging eGFRs and serum albumin concentrations, approximately 67 years of age. The main objectives were to assess flucloxacillin population PK and determine a dosing strategy that maximizes PK/PD target attainment in critically ill patients based on dosing simulations.
Discussion
In the present study, we describe the development of a population PK model for flucloxacillin in critically ill patients and consecutively PK/PD target attainment in this population, based on dosing simulations. The main study finding was that critically ill patients were at a considerable risk of underdosing when flucloxacillin was employed in standard daily doses of up to 12 g.
Dosing simulations revealed only 26% PK/PD target attainment (≥ 50% ƒT
>2 mg/L) following daily continuous infusion of 12 g flucloxacillin. These results are inconsistent with findings of previous dosing simulation studies performed in critically ill patients [
10,
11,
20]. Two studies reported ≥ 99.9% target attainment for 8 to 12 g per 24 h, with target serum concentrations of 2 to 2.5 mg/L [
11,
20]. In addition, Jager et al
. have reported 91% target attainment in patients with an eGFR of 33 mL/min and 71% for an eGFR of 153 mL/min with 2 g administered 6 times daily (q4h), accompanied by a PK/PD target of 100% ƒT
>0.5 mg/L [
10]. However, our study results are in line with PK/PD target attainment as reported in two prospective, observational studies [
19,
21]. Moser et al
. have reported 26% target attainment (100% ƒT
>2 mg/L) for 2 g administered 4 to 6 times daily; however, the authors also reported ‘optimal’ PK/PD target attainment of 90% when target serum concentrations were based on strain-specific MICs or 0.25 mg/L [
19]. Wong et al
. [
21] have documented 52% target attainment (100% ƒT
>strain-specific MIC) for 2 g q4h and 30% target attainment for 100% ƒT
>4 x strain-specific MIC.
Several aspects could have contributed to differences in percentages of flucloxacillin PK/PD target attainment between our study and those reported previously [
10,
11,
19‐
21], including (1) heterogeneity of the critically ill population, (2) complexity of plasma protein binding, and (3) appropriate selection of the target serum flucloxacillin concentration.
First, critically ill patients are known to exhibit considerable heterogeneity [
7]. Previous studies have focused on critically ill subpopulations, complicating the comparison of study results [
10,
20,
21]. Wallenburg et al
. [
11] have performed a dosing simulation study in a population most comparable to the present study population. Despite the older age of our study population, the calculated eGFR was comparable between both studies. In addition, non-renal drug clearance is generally preserved in elderly patients [
36]. However, we detected a substantially reduced PK/PD target attainment, which may partly be explained by an elevated median flucloxacillin clearance of 77.5 L/h when compared with 37.5 L/h. In addition, the study population of Wallenburg et al
. [
11] consisted of 21% of patients who underwent continuous RRT and patients with liver cirrhosis may have been included, whereas these patients were excluded in our study. Furthermore, we noted a slightly elevated ƒ
u, potentially resulting in increased non-renal clearance and tubular secretion. The PK model performance was improved by incorporating albumin and eGFR covariates, which is consistent with previous study results [
10,
11]. No other significant model covariates were found to alter flucloxacillin PK, protein binding, and clearance. However, our study population presented a high body weight and BMI, along with an increased volume of distribution, and consisted of older patients [
10,
11,
20]. These aspects may have contributed to the remaining proportional PK model error of up to 42%.
Second, plasma protein binding of flucloxacillin in critically ill patients remains complex [
37‐
39]. Flucloxacillin and albumin concentrations reportedly impact protein binding and PK [
10,
11,
40,
41]. However, these individual values may be difficult to interpret, for instance, due to both covalent and non-covalent bindings of flucloxacillin to plasma proteins or penicillin-induced pseudo-hypoalbuminemia [
42‐
45]. The median observed ƒ
u in our study was 22%, which was slightly higher than the 7 to 19% reported in previous studies [
10,
11,
19]. The observed broad ƒ
u range of 6–73% in our study is in line with previous studies [
10,
11,
19]. Interestingly, the median serum albumin concentration in the present study was slightly higher than in most previous studies [
10,
11,
19,
20]. However, the higher ƒ
u might be related to our older study population. For instance, plasma protein binding and flucloxacillin displacement from plasma proteins could be altered in older ICU patients owing to endogen molecules and polypharmacy [
40,
46]. Inter-individual variance (IIV) on albumin or ƒ
u in our study was higher than that reported in other studies [
10,
11,
19‐
21], mainly related to the exclusion of patients with serum albumin concentrations > 32 g/L in several previous studies [
10,
20].
Third, target unbound serum flucloxacillin concentrations remain poorly defined [
6,
47,
48]. An ECOFF value for flucloxacillin is lacking [
30] but is stated to be similar to that of oxacillin and cloxacillin. However, cloxacillin ECOFF is 0.5 mg/L, and oxacillin ECOFF is 2 mg/L [
30]. In the present study, we selected a target of 50% ƒT
>4x0.5 mg/L, representing a target serum concentration of 2 mg/L; if a MIC of 2 mg/L had been selected, we would have attained even lower target attainment percentages. In addition, some studies mentioned target concentration selection based on strain-specific MICs from positive blood cultures [
19,
21]. However, target concentration selection based on a single MIC determination has been deemed inappropriate and could be detrimental to patient therapy [
48,
49]. First, routine clinical laboratories cannot accurately determine individual MICs owing to the inherent assay variation. Second, biological variation exists within a species even when there are no acquired resistance mechanisms [
48]. Furthermore, we selected a serum target concentration of 4 times the ECOFF value [
30]. Some previous studies reported improved clinical or microbiological cure for beta-lactam antibiotics when serum concentrations 2.1 to 5 times the MIC were achieved [
12,
13,
15]. In critically ill patients, a higher incidence of more resistant pathogens is reported, and antibiotic tissue penetration may be impaired [
2,
4,
13‐
18]. Therefore, to optimize antimicrobial efficacy and battle antimicrobial resistance, it is essential to eliminate all targeted pathogens, and target concentrations should be based on ECOFF values [
30,
48,
49].
Our study has certain limitations. First, limited sampling of flucloxacillin (1–3 samples) was performed for most patients (74%), which could have potentially resulted in a suboptimal description of the individual PK and, consequently, the predicted population PK. Conversely, rich sampling data were available from 8 patients (26%), with up to 28 samples per patient. Additionally, a multicenter study was performed, and appropriate population PK model performance was demonstrated. Second, dosing simulations were performed; however, the collection of flucloxacillin concentration measurements from real patients would have been preferred. Unfortunately, the inclusion of large numbers of critically ill patients in PK studies can be challenging [
33]. Third, patients suffering from liver cirrhosis or receiving RRT were excluded from the present study, restricting the current study results from representing the entire ICU population. Fourth, CL
cr was not actually measured, but we used the CKD-EPI equation to estimate CL
cr. More accurate methods have been described in critically ill patients, such as calculating urine-to-plasma creatinine ratios [
50]. Fifth, individual pathogen and MIC determination were not acquired, which could have aided the interpretation of appropriate flucloxacillin exposure.
Future research should focus on identifying efficacy and toxicity thresholds to maximize antimicrobial exposure and efficacy in critically ill patients [
21]. Also, further research is required to assess which specific patients are at risk for flucloxacillin underexposure. For instance, underexposure may be related to age, APACHE II score or time since flucloxacillin treatment initiation. In addition, to comprehensively elucidate the in vivo equilibrium between protein-bound and unbound flucloxacillin, the complexity of flucloxacillin plasma protein binding needs to be further unraveled.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.