Population Pharmacokinetics of Fosfomycin in Critically Ill Patients

Abstract

This study describes the population pharmacokinetics of fosfomycin in critically ill patients. In this observational study, serial blood samples were taken over several dosing intervals of intravenous fosfomycin treatment. Blood samples were analyzed using a validated liquid chromatography-tandem mass spectrometry technique. A population pharmacokinetic analysis was performed using nonlinear mixed-effects modeling. Five hundred fifteen blood samples were collected over one to six dosing intervals from 12 patients. The mean (standard deviation) age was 62 (17) years, 67% of patients were male, and creatinine clearance (CL_CR) ranged from 30 to 300 ml/min. A two-compartment model with between-subject variability on clearance and volume of distribution of the central compartment (V_c) described the data adequately. Calculated CL_CR was supported as a covariate on fosfomycin clearance. The mean parameter estimates for clearance on the first day were 2.06 liters/h, V_c of 27.2 liters, intercompartmental clearance of 19.8 liters/h, and volume of the peripheral compartment of 22.3 liters. We found significant pharmacokinetic variability for fosfomycin in this heterogeneous patient sample, which may be explained somewhat by the observed variations in renal function.

INTRODUCTION

Inadequate treatment of infections among patients requiring intensive care unit (ICU) admission is an important determinant of hospital mortality (1). Drug dosing should be considered an essential part of optimizing antibiotic use. However, critically ill patients have been shown to have significant pharmacokinetic variability for some antibiotics due to the physiological changes associated with this pathology (2). This variability has been shown to affect the achievement of therapeutic exposures of antibiotics (3–7). If not considered in dosing regimens, this pharmacokinetic variability can lead to clinical failure or toxicity (8).

Fosfomycin is a broad-spectrum antibiotic with bactericidal activity against various Gram-negative and Gram-positive bacteria and has been gaining considerable attention recently due to its effectiveness as a treatment for multidrug-resistant pathogens (9), including extended-spectrum beta-lactamase- and carbapenemase-producing bacteria. Fosfomycin exhibits extensive penetration into many tissue types (10–13) and is well tolerated, with only minor adverse events reported (14, 15). However, there are few data on the pharmacokinetics of this potentially valuable antibiotic in critically ill patients, and therefore, significant uncertainty relating to appropriate dosing exists.

Dosing guidelines from the European Committee on Antimicrobial Susceptibility and Testing (EUCAST) recommend dosing schedules of 3 to 4 g intravenous (i.v.) fosfomycin three times daily up to a maximum dose of 5 to 8 g fosfomycin three times daily (16). However, the data supporting these recommendations are unclear given that we are unaware of any studies investigating the population pharmacokinetics of i.v. fosfomycin in critically ill patients.

The aim of this study was to describe the population pharmacokinetics of fosfomycin in critically ill patients.

MATERIALS AND METHODS

Patients.

This study was performed in the ICUs of two hospitals (Attikon University Hospital and Hygeia Hospital, Athens, Greece). Ethical approval to conduct the study was obtained from the local institutional ethics committees. Consent to participate was obtained from the patient or the patient's legally authorized representative.

Critically ill patients who were prescribed i.v. fosfomycin by the treating physician were eligible for inclusion. In accordance with the usual practice, all patients had an indwelling arterial cannula. Patients meeting any of the following criteria were excluded: (i) age less than 18 years, (ii) use of fosfomycin within the previous month, (iii) pregnancy or lactation, and (iv) consent not obtained.

The severity of illness of each patient was described using the acute physiology and chronic health evaluation II (APACHE II) (17) and sequential organ failure assessment (SOFA) (18) scores on days 1 and 4 of treatment. Serum creatinine concentrations were collected as a routine procedure in all patients, with creatinine clearance (CL_CR) calculated daily using the Cockcroft-Gault equation (19). A measured urinary CL_CR over a 24-h time period was collected on the first day of sampling. Serum biochemistry values, including albumin concentration, white blood cell count, and bilirubin concentration, were recorded. Concomitant antibiotic treatment was also recorded, as was length of stay in the ICU and hospital, overall mortality, and outcome of treatment.

Fosfomycin treatment.

In accordance with the study protocol, treatment with fosfomycin was added to therapy only after the culture results became available. The fosfomycin MIC was determined by Etest (bioMérieux, Marcy l'Etoile, France). The results were interpreted according to EUCAST 2014 criteria (20). Administration of fosfomycin was by i.v. infusion over 30 to 60 min in accordance with local guidelines. The choice of antibiotic dose was at the discretion of the treating physician, with doses of 4 or 6 g of fosfomycin at a frequency of three or four times daily being common in both ICUs.

Sample collection.

Blood samples were collected from an indwelling arterial cannula before the drug administration and at 30 min, 45 min, 1 h, 1.5 h, 2 h, 4 h, and 6 h after administration of fosfomycin. Where possible, sampling occurred during the first dosing interval and/or on days 2, 4, 5, 6, and 7.

Drug assay.

Plasma fosfomycin concentrations were determined using a previously described high-performance liquid chromatography-tandem mass spectrometry method (21) at the Burns Trauma and Critical Care Research Centre, The University of Queensland, Australia. The assay interday coefficients of variation for fosfomycin in plasma were ≤9.1%, with an accuracy range of −7.2 to 3.3%. The assay limit for plasma was 1 mg/liter, with precision at 3.6% and accuracy of 1.1%. The linearity of the assay (r²) was 0.9963 (n = 12).

Pharmacokinetic and statistical analysis.

The concentration-time data were analyzed using nonlinear mixed-effects modeling (NONMEM version 7.3; Globomax LLC, Hanover, MD, USA). A digital Fortran compiler was used, and the runs were executed using Wings for NONMEM (http://wfn.sourceforge.net). The first-order conditional estimation method with interaction was used throughout the model building.

Model development.

For the population pharmacokinetic analysis, the one- and two-compartment linear models were fitted into plasma fosfomycin concentration data, using subroutines from the NONMEM library. Between-subject variability (BSV) was best described using an exponential variability model. Residual unexplained variability was tested using various model iterations.

Model diagnostics.

The goodness of fit of the model was evaluated using visual inspection of diagnostic scatter plots and the NONMEM objective function value (OFV). A statistical comparison of nested models was undertaken, with a decrease in OFV of 3.84 units (P < 0.05) considered statistically significant.

Covariate screening.

Covariate model building was performed using sequential assessment of biologically plausible clinical parameters. Forward inclusion was based upon the aforementioned model selection criteria and significant correlation with one of the pharmacokinetic parameters. The covariates evaluated were the calculated and measured urinary CL_CR, age, sex, weight, SOFA and APACHE scores, and serum albumin concentration.

Bootstrap.

A NONMEM nonparametric bootstrap method (n = 1,000) was used to describe the uncertainty of the pharmacokinetic parameter estimates in the final model. Using the bootstrap empirical posterior distribution, we obtained the 95% confidence interval (CI; 2.5th to 97.5th percentiles) for the model parameters, using previously described methods (22).

RESULTS

Patient characteristics.

A total of 515 plasma samples were collected over one to six dosing intervals from 12 enrolled patients. The demographic and clinical characteristics of the patients are shown in Table 1. All patients received a dose of 6 g of fosfomycin every 6 h, except for patients 7 and 9. Patient 7 received a dose of 4 g of fosfomycin every 6 h, and patient 9 received a dose of 6 g of fosfomycin every 8 h. The microbiology for these infections, concomitant antibiotics, and patient outcomes are described in Table 2. All patients were diagnosed as having septic shock and respiratory failure, and all patients were intubated.

TABLE 1.

Patient characteristics

Characteristica	Value [median (IQRb) or as indicated]
Age (yr)	62.5 (57.8 to 75.0)
Weight (kg)	71.5 (69.5 to 80.0)
Body mass index (kg/m2)	26.4 (23.9 to 27.5)
No. (%) male/no. (%) female	8 (67)/4 (33)
APACHE II score on ICU admission	11.5 (8.8 to 16.5)
SOFA score on ICU admission	7 (6 to 10)
CLCR on admission (ml/min)	59 (52 to 99)
Albumin concn (g/dl)	2.7 (2.5 to 3.0)

^aAPACHE, acute physiology and chronic health evaluation; ICU, intensive care unit; SOFA, sequential organ failure assessment; CL_CR, creatinine clearance.

^bIQR, interquartile range.

TABLE 2.

Specimen types, organisms isolated and their susceptibilities, concomitant antibiotics, lengths of stay, and treatment outcomes

Patient	Specimen type	Organism isolateda	Drug(s) used (susceptibility; MIC [mg/liter])b	Concomitant antibiotic(s)	Length of stay (days)		Treatment outcome
					ICUc	Hospital
1	Tracheal aspirate	KPC Klebsiella pneumoniae	Gentamicin (I), fosfomycin (S; 24)	Gentamicin, tigecycline	42	60	Improvement
2	Peritoneal fluid	KPC Klebsiella pneumoniae	Gentamicin (S)	Gentamicin, colistin, linezolid	28	90	Improvement
	Bronchial secretion	Pseudomonas aeruginosa	Colistin (S)
		Acinetobacter baumannii	Colistin (S)
		KPC Klebsiella pneumoniae	Gentamicin (S), fosfomycin (S; 18)
	Stool	KPC Klebsiella pneumoniae	Amikacin (I), fosfomycin (S; 24)
3	Tracheal aspirate	Klebsiella pneumoniae	PDR; fosfomycin (R; 128)	Ampicillin-sulbactam, colistin, cefepime	28	48	Death on 8th day of treatment
		Acinetobacter baumannii	PDR
		Stenotrophonomas maltophilia	PDR
4	Bronchial aspirate	KPC Klebsiella pneumoniae	Cefepime (I), gentamicin (I), meropenem (I), fosfomycin (S; 32)	Cefepime	46	66	Death on 9th day of treatment
		Acinetobacter baumannii	Ampicillin-sulbactam (I)
		Candida parapsilosis
	Urine	Candida parapsilosis
5	Pleural effusion	Stenotrophonomas maltophilia	Minocycline (S), trimethoprim-sulfamethoxazole (S), levofloxacin (S),	Meropenem, linezolid	42	82	Death 11 days after end of treatment
		KPC Klebsiella pneumoniae	Gentamicin (S), tobramycin (S), amoxicillin (S), colistin (S), fosfomycin (S; 32)
6	Tracheal aspirate	Acinetobacter baumannii	PDR	Gentamicin, meropenem, colistin	40	55	Death on 2nd day of treatment
	Stool	KPC Klebsiella pneumoniae	Gentamicin (S), fosfomycin (S; 42)
7	Surgical wound of septic arthritis	Acinetobacter baumannii	Colistin (S)	Tigecycline, meropenem, colistin	53	76	Death on 15th day of treatment
		Pseudomonas aeruginosa	Colistin (S), aztreonam (S)
	Bone from septic arthritis	Klebsiella pneumoniae	Gentamicin (S), trimethoprim-sulfamethoxazole (S), fosfomycin (S; 32)
	Urine	KPC Klebsiella pneumoniae	PDR; fosfomycin (R; 128)
	Rectal	KPC Klebsiella pneumoniae	PDR; fosfomycin (S; 32)
		Acinetobacter baumannii	Colistin (S)
8	Sputum	KPC Klebsiella pneumoniae	Amikacin (S), minocycline (S), tetracycline (S), tigecycline (S), colistin (S), fosfomycin (S; 32)	Colistin, tigecycline	12	35	Death on 4th day of treatment
		Pseudomonas aeruginosa	Colistin (S), fosfomycin (R)
9	Bronchial secretion	KPC Klebsiella pneumoniae	Fosfomycin (S; 64)		25	25	Death on 4th day of treatment
10	Blood	Klebsiella pneumoniae	Fosfomycin (S; 32)	Colistin	36	36	Improvement
11	Bronchial secretion	Klebsiella pneumoniae	Fosfomycin (S; 18)	Meropenem	14	14	Improvementd
12	Blood	Klebsiella pneumoniae	Colistin (S), fosfomycin (S; 64)	Colistin	6	210	Improvemente

^aKPC, Klebsiella pneumoniae carbapenemase-producing organism.

^bI, intermediate; S, susceptible; R, resistant; PDR, pandrug resistant.

^cICU, intensive care unit.

^dFosfomycin was discontinued after 1 day due to infection identified as due to extended-spectrum beta-lactamase-producing Klebsiella pneumoniae.

^eFosfomycin was discontinued after 1 day due to allergic rash.

Patient plasma-concentration data.

The median trough fosfomycin plasma concentration (C_min) for patients during the first sampling interval was 84.3 mg/liter (range, 41 to 172 mg/liter). On day 2, the median C_min was 250 mg/liter (range, 76 to 684 mg/liter). On day 7, two patients had a C_min of >1,000 mg/liter, while two others had a C_min of <100 mg/liter.

Pharmacokinetic analysis.

The best base model consisted of a two-compartment linear model with zero-order input (ADVAN3 TRANS4) incorporating an exponential residual unknown variability. BSV was supported on clearance (CL) and volume of the central compartment (V_c). The inclusion of between-subject variability for CL and V_c was associated with decreases in the objective functions of 659 and 64.4, respectively.

The calculated CL_CR using the Cockcroft-Gault equation on CL was associated with a decrease in the objective function of 79.7 and was supported as a covariate in the final model. Inclusion of allometrically scaled patient weight for the V_c improved the goodness-of-fit plots and was therefore also included in the model.

The final model is represented as follows: TV_CL = (θ_1–6 × CL_CR/90) and TV_Vc = θ × (WT/70)^0.75, where TV_CL is the typical value of CL; θ_1–6 is the typical value of fosfomycin CL in the population, with each sampling day defined as an individual occasion (θ), from days 1 (θ₁), 2 (θ₂), 4 (θ₃), 5 (θ₄), 6 (θ₅), and 7 (θ₆); CL_CR is normalized to the mean value for the patients; WT is patient weight; and TV_Vc is the typical value of V_c.

Figure 1a displays the diagnostic goodness-of-fit plot for the final population pharmacokinetic covariate model. Figure 1b displays a visual predictive check for the final covariate model from the first day of sampling through to the sixth day of sampling. The diagnostic plots in Fig. 1 demonstrate that the final pharmacokinetic model provides an adequate description of the measured concentrations of fosfomycin from this highly heterogenous sample of patients of widely varied age and renal function. The fit of the model was acceptable in terms of visual or statistical biases for the prediction. The parameter estimates for the final model are given in Table 3 and include the 95% values from all bootstrap runs. Having met the criteria for inclusion, the model was accepted.

FIG 1.

Diagnostic plots for the final population pharmacokinetic covariate model. (a) Individual predicted fosfomycin concentrations versus observed concentrations (r² = 0.97); the nonlinear regression line of best fit is shown by the continuous line, and the line of identity xy is shown by the dotted line. (b) Visual predictive check generated from a Monte Carlo simulation (n = 1,000) showing that the estimated population pharmacokinetic model has adequate performance (data from 0 to 150 h of dosing only are shown); the raw data for individual patients are shown as dots, the continuous lines represent the mean estimates, and the dotted lines represent the 95% confidence interval limits.

TABLE 3.

Bootstrap parameter estimates of the final covariate model

Parameter	Median value for model	Value for bootstrap estimate
		Median	95% CI
			2.5th percentile	97.5th percentile
Fixed effects
Clearance (liters/h) on occasion:
1	2.06	2.19	1.11	5.11
2	5.57	5.06	2.76	9.37
4	4.67	4.05	2.46	7.42
5	4.13	4.13	2.29	9.37
6	5.42	4.93	2.59	9.96
7	5.19	4.74	2.31	9.09
Volume (liters) of compartment
Central	26.5	26.4	19.0	34.2
Peripheral	22.3	21.6	12.7	34.9
Intercompartmental CL (liters/h)	19.8	19.8	8.01	46.0
Random effects, BSV (%CV)a
Clearance (liters/h)	91.9	81.9	48.1	128
Volume of central compartment (liters)	39.0	36.7	18.0	57.2
Random error, exponential (%CV)	16.4	15.5	10.5	19.2

^aBSV, between-subject variability; CV, coefficient of variation.

DISCUSSION

Fosfomycin is currently being used as a last-line treatment for critically ill patients for the treatment of serious infections, yet there is limited information available describing the pharmacokinetics of fosfomycin in this subpopulation (9). We found that the CL was proportional to the calculated CL_CR, with increasing CL_CR values affecting the likelihood that fosfomycin doses would achieve target concentrations for relevant MICs.

A previous study from W. M. M. Kirby tested i.v. doses of fosfomycin of 4 g every 6 h in healthy patients and showed accumulation, with peak serum concentrations increasing from 195 mg/liter after the first dose up to 253 mg/liter after the second dose (23). CL was reported as 7.2 liters/h. This mean CL is substantially higher than our mean CL of 2.06 liters/h (first day of patient sampling of our study) and demonstrates the need for careful monitoring of fosfomycin dosing in critically ill patients, particularly if there is evidence of renal dysfunction. While fosfomycin is well tolerated and has a low adverse event profile (14), we observed peak plasma concentrations of up to 1,440 mg/liter after multiple days of treatment in some patients with low CL_CR values.

The mean apparent volume of distribution in this study was 48.8 liters, which is higher than that observed by Kirby (22 liters) (23) in healthy volunteers, likely due to the presence of more severe illness. A larger than normal volume of distribution is one of the typical pathophysiological changes observed in critically ill patients and has been reported in multiple pharmacokinetic studies for different antibiotics (7, 24).

In this study of 12 patients, two patients consistently had CL_CR values of >200 ml/min. Augmented renal clearance—defined as a sustained elevation of CL_CR (>130 ml/min/1.73 m²)—manifests in over 65% of critically ill patients in the first week of admission into an ICU (25) and, as seen in this study, is likely to affect the probability of attainment of therapeutic exposures of renally cleared antibiotics. Renal insufficiency is associated with dramatically higher fosfomycin concentrations, and two patients consistently had CL_CR values of <50 ml/min/1.73 m². The pharmacokinetic variability observed in this study is explained somewhat by the wide ranges of renal function.

This study has some limitations we would like to declare. We did not measure free concentrations of fosfomycin in plasma or the concentrations at the site of infection. Instead, we measured the total drug concentration. However, the antimicrobial activity of fosfomycin has been shown to not be affected by the presence of albumin (26) and is considered to have negligible protein binding. Also, given the high extremes of renal function in patients and sample collection over up to 6 dosing intervals, the fact that this study is of a relatively small cohort of 12 patients could also be considered a limitation. Indeed the small, heterogeneous sample may also affect the ability of the model to identify other relevant covariates, although this number of patients provides useful data that clinicians can use to procure more relevant doses (27). The visual predictive check (Fig. 1b) suggests that there is a small overestimation of higher concentrations from the model compared to the individual data points in the later days of sampling. However, the data in Fig. 1a show that the line of x = y overlaps with the regression line, confirming that it is not a systematic deviation, which supports the adequacy of the final model. Finally, fosfomycin was coadministered as part of a more extensive antibiotic treatment in 8 of the 12 patients, and this model was not intended to provide any description of pharmacodynamic synergistic effects.

Conclusion.

This population pharmacokinetic study found a substantially lower mean CL and larger than normal volume of distribution of fosfomycin compared to those previously reported in healthy subjects. The mean parameter estimates for clearance on the first day were 2.06 liters/h, V_c of 27.2 liters, intercompartmental clearance of 19.8 liters/h, and volume of the peripheral compartment of 22.3 liters. We found significant pharmacokinetic variability of fosfomycin in this heterogeneous patient sample, which may be explained in part by the variations observed in renal function.