MAJOR ARTICLE

Population Pharmacokinetics of Intravenous

Polymyxin B in Critically Ill Patients:
Implications for Selection of Dosage Regimens
Ana M. Sandri,1,a Cornelia B. Landersdorfer,2,3,a Jovan Jacob,4 Márcio M. Boniatti,5 Micheline G. Dalarosa,6
Diego R. Falci,6 Tainá F. Behle,7 Rosaura C. Bordinhão,6 Jiping Wang,4 Alan Forrest,3 Roger L. Nation,4 Jian Li,4,b and
Alexandre P. Zavascki7,b
1
Infectious Diseases Service, Hospital São Lucas da Pontifícia Universidade Católica do Rio Grande do Sul, Porto Alegre, Brazil; 2Centre for Medicine Use
and Safety, Faculty of Pharmacy and Pharmaceutical Sciences, Monash University, Parkville, Victoria, Australia; 3School of Pharmacy and Pharmaceutical
Sciences, University at Buffalo, State University of New York; 4Drug Delivery, Disposition and Dynamics, Monash Institute of Pharmaceutical Sciences,
Monash University, Parkville, Victoria, Australia; 5Intensive Care Unit, Hospital de Clínicas de Porto Alegre, 6Infection Control Service, Hospital Nossa

Downloaded from http://cid.oxfordjournals.org/ by guest on January 9, 2016

Senhora da Conceição, and 7Infectious Diseases Service, Hospital de Clínicas de Porto Alegre, Porto Alegre, Brazil

Background. Polymyxin B is a last-line therapy for multidrug-resistant gram-negative bacteria. There is a

dearth of pharmacokinetic data to guide dosing in critically ill patients.
Methods. Twenty-four critically ill patients were enrolled and blood/urine samples were collected over a dosing
interval at steady state. Polymyxin B concentrations were measured by liquid chromatography–tandem mass spec-
trometry. Population pharmacokinetic analysis and Monte Carlo simulations were conducted.
Results. Twenty-four patients aged 21–87 years received intravenous polymyxin B (0.45–3.38 mg/kg/day). Two
patients were on continuous hemodialysis, and creatinine clearance in the other patients was 10–143 mL/min. Even
with very diverse demographics, the total body clearance of polymyxin B when scaled by total body weight ( popula-
tion mean, 0.0276 L/hour/kg) showed remarkably low interindividual variability (32.4% coefficient of variation).
Polymyxin B was predominantly nonrenally cleared with median urinary recovery of 4.04%. Polymyxin B total
body clearance did not show any relationship with creatinine clearance (r 2 = 0.008), APACHE II score, or age.
Median unbound fraction in plasma was 0.42. Monte Carlo simulations revealed the importance of initiating thera-
peutic regimens with a loading dose.
Conclusions. Our study showed that doses of intravenous polymyxin B are best scaled by total body weight.
Importantly, dosage selection of this drug should not be based on renal function.
Keywords. polymyxins; pharmacokinetics; dose selection; plasma protein binding; urinary recovery.

The burgeoning multidrug resistance among gram- There are 2 polymyxins used clinically, polymyxin B and
negative bacteria, combined with a paucity of new anti- colistin (ie, polymyxin E) [1, 2]. Both antibiotics were
biotics, has led to the reemergence of polymyxins [1, 2]. first used clinically in the late 1950s but were largely
abandoned in the 1970s due to toxicity; however, they
were reintroduced to the therapeutic arsenal in the last
Received 26 January 2013; accepted 7 May 2013; electronically published 22 decade due to resistance to all other antibiotics [1–4].
May 2013. Pharmacokinetics (PK)/pharmacodynamics (PD) of
a
A. M. S. and C. B. L. contributed equally to this study.
b
J. L. and A. P. Z. were joint senior authors and contributed equally to this study. antibiotics is critical for optimizing their dosage regi-
Correspondence: Jian Li, PhD, Drug Delivery, Disposition and Dynamics, Monash mens to maximize efficacy and minimize toxicity and
Institute of Pharmaceutical Sciences, Monash University, Parkville, Victoria, Austra-
lia ( jian.li@monash.edu).
resistance. Almost all modern PK studies on polymyx-
Clinical Infectious Diseases 2013;57(4):524–31 ins are for colistin that is administered parenterally as
© The Author 2013. Published by Oxford University Press on behalf of the Infectious its inactive prodrug, colistin methanesulfonate (CMS) [5].
Diseases Society of America. All rights reserved. For Permissions, please e-mail:
journals.permissions@oup.com.
In contrast, polymyxin B is available for direct parenter-
DOI: 10.1093/cid/cit334 al administration, that is, as the antibacterial entity [2].

524 • CID 2013:57 (15 August) • Sandri et al

Recent PK studies have highlighted that the conversion of CMS infusion and immediately before the next infusion. Blood sam-
to colistin occurs very slowly and incompletely in vivo [6, 7]. ples were centrifuged for 10 minutes (4000 g), and the plas-
Therefore, current PK findings for CMS/colistin cannot be ex- ma samples were immediately stored at −80°C until analysis.
trapolated to polymyxin B. Urine was collected across the dosage interval (0–6 hours and
There are no scientifically based dosing guidelines for poly- 6–12 hours for 12-hourly dosing; 0–6 hours, 6–12 hours, and
myxin B due to the lack of solid PK information. Reports on 12–24 hours for 24-hourly dosing) and the volumes were re-
the PK of polymyxin B in humans have been limited to a small corded; 5 mL of each collection was stored at −80°C pending
number of studies with 20 patients in total [8–11]. Even though analysis.
a population PK study was reported in 9 adult patients [9], only
2 serum samples were collected from each patient and, unfortu- Binding of Polymyxin B in Plasma
nately, only polymyxin B1 was measured, an approach leading Plasma protein binding of polymyxin B was determined by
to the potential for bias in defining polymyxin B PK. Further- rapid equilibrium dialysis [11]. For each patient, 3 plasma
more, no urine samples were collected in that study and, therefore, samples were pooled ( pH adjusted to 7.4) and 1 mL was dia-
urinary recovery and renal clearance of polymyxin B could not be lyzed against an equal volume of isotonic phosphate-buffered
obtained [9]. Importantly, the applicability of that study [9] to saline ( pH 7.4) at 37°C for 4 hours. Polymyxin B concentra-
critically ill patients is limited. Hence, there is an urgent need to tions were determined (described below), and unbound frac-
investigate the PK of polymyxin B in critically ill patients and tion in plasma (fu) was calculated as the ratio of the
to optimize its clinical use. In this study, we developed a popu- concentration in buffer to that in plasma.

Downloaded from http://cid.oxfordjournals.org/ by guest on January 9, 2016

lation PK model for polymyxin B after intravenous administra-
tion in 24 patients with various degrees of renal function. Our Quantification of Polymyxin B Concentrations
aim was to identify patient factors influencing the PK and thereby Polymyxin B (base) concentrations in samples were quantified
allow proposing of dosage regimens to achieve a desired target by monitoring both polymyxin B1 and B2 using a validated
plasma polymyxin B concentration. The study also provides ultraperformance liquid chromatography–tandem mass spec-
important new information on the renal handling of polymyxin trometry (LC-MS/MS) assay [11]. Analysis of independently
B that assists in understanding the potential for this antibiotic prepared quality control samples indicated good reproducibility
to cause nephrotoxicity. (coefficients of variation ≤8.39%) and accuracy (measured con-
centrations ≤10.0% from target concentrations). The limit of
PATIENTS AND METHODS quantification was 0.05 mg/L.

Patients and Ethics Pharmacokinetic Analysis

The study was approved by the Ethical Committees of Hospital Nonlinear mixed-effects modeling of polymyxin B population
de Clínicas de Porto Alegre, Hospital São Lucas da Pontifícia PK was performed utilizing the S-ADAPT platform (version
Universidade Católica do Rio Grande do Sul, Hospital Nossa 1.57) with the Monte Carlo parametric expectation maximiza-
Senhora da Conceição, Brazil, and Monash University, Austra- tion algorithm [12]. The SADAPT-TRAN program was used for
lia. Informed consent was obtained from all patients or their pre- and postprocessing [13, 14]. One-, 2-, and 3-compartment
legal representatives. The decision to administer polymyxin B models were explored for the polymyxin B plasma concen-
and its dosing regimen was made by the attending physician. tration-time profiles with linear or nonlinear (saturable) eli-
From March 2011 to January 2012, 24 patients (aged ≥18 mination. These models were fit to the data for all patients
years) who received intravenous polymyxin B were included; simultaneously. The interindividual variability was assumed to
the noncompartmental PK data of 2 patients on continuous ve- be log-normally distributed. Candidate covariates that were ex-
novenous hemodialysis (CVVHD) have been reported [11]. plored for their possible effect on polymyxin B disposition in-
cluded the following: body size (total body weight [TBW] and
Polymyxin B Administration and Sample Collection lean body weight [LBW]), sex, age, creatinine clearance (CrCL),
Blood samples were collected after ≥48 hours of treatment with serum albumin concentration, and APACHE II score on total
polymyxin B (sulfate; Polymyxin B for Injection, Eurofarma, body clearance (CL) of polymyxin B; and body size (TBW and
Brazil). Collection of urine samples was possible for 17 of the LBW) on volume of distribution. CrCL was calculated using
patients. Polymyxin B was administered by short-term infu- the Cockcroft-Gault equation with LBW [15]. For model evalu-
sions (60–240 minutes) every 12 hours (23 patients) or every ation, plots of observed versus individual-fitted and observed
24 hours (1 patient). In each patient, 8 blood samples (3 mL versus population-fitted polymyxin B concentrations, the nor-
each) were collected immediately before starting the infusion, 5 malized prediction distribution error, and the objective func-
minutes, and 0.5, 1, 2, 4, and 8 hours after completing the tion in S-ADAPT were utilized. Polymyxin B renal clearance

Pharmacokinetics of Polymyxin B • CID 2013:57 (15 August) • 525

(CLR) was calculated as the amount recovered in urine during Table 1. Patient Characteristics
the urine collection period divided by the area under the plasma
concentration-time curve (AUC) across the same period. Because Characteristic Valuea
only unbound drug in plasma is filtered at the glomeruli, the Age, y 61.5 (21–87)
clearance of polymyxin B by glomerular filtration (CLGF) was es- Sex
timated as fu × GFR (where GFR is glomerular filtration rate Male 13 (54.2)
from CrCL) [16]. The percentage of polymyxin B reabsorbed Female 11 (45.8)
from tubular urine was calculated as 100 × (1–CLR/CLGF); the Total body weight, kg 62.5 (41–250)
percentage of water reabsorbed was similarly determined as Lean body weight, kg 46.0 (29–99)
Estimated creatinine clearance (mL/min) 33 (10–143)
100 × (1–[urine flow rate]/GFR). The number of milligrams per
APACHE II score 21.5 (10–29)
day of polymyxin B filtered at the glomeruli was calculated by
Coadministered antibiotic(s) 24 (100)
multiplying the polymyxin B average steady-state plasma con-
Vancomycin 18 (75)
centration (Css,avg, mg/L) with its filtration clearance (CLGF, L/ Trimethoprim-sulfamethoxazole 6 (25)
day); the milligrams of polymyxin B per day reabsorbed by the Levofloxacin 2 (8.3)
renal tubular cells was then estimated from the percentage of the Meropenem 4 (16.7)
filtered polymyxin B that was reabsorbed. Ceftazidime 3 (12.5)
Cefepime 1 (4.2)
Piperacillin-tazobactam 1 (4.2)
Monte Carlo Simulations of Dosage Regimens

Downloaded from http://cid.oxfordjournals.org/ by guest on January 9, 2016

With renal replacement therapy
Simulations with between-subject variability (BSV) were per-
Continuous renal replacement 2 (8.3)
formed for various clinically relevant dosage regimens scaled
Abbreviation: APACHE II, Acute Physiology and Chronic Health Evaluation II.
by TBW. The regimens were as follows: (1) 1.25 mg/kg as a 1- a
Values are median (range) or No. (%).
hour infusion every 12 hours; (2) 2 mg/kg as a 2-hour infusion
loading dose followed 12 hours later by 1.25 mg/kg as a 1-hour
infusion every 12 hours; (3) 1.5 mg/kg as a 1-hour infusion
every 12 hours; (4) 2.5 mg/kg as a 2-hour infusion loading dose The time course of polymyxin B concentrations was well de-
followed 12 hours later by 1.5 mg/kg as a 1-hour infusion every scribed by a 2-compartment disposition model with constant-
12 hours; (5) continuous infusion of 2.5 mg/kg/24 hours; and rate, short-term infusion input and linear (ie, nonsaturable)
(6) 2 mg/kg as a 2-hour infusion loading dose immediately fol- elimination. The population PK model provided excellent fits
lowed by continuous infusion of 2.5 mg/kg/24 hours. For each to the observed concentration-time profiles for individual pa-
dosage regimen, 5000 virtual subjects were simulated using tients (Figure 2), and the individual-fitted and population-
NONMEM software (version VI, level 1.2) [17]. fitted concentrations were unbiased and adequately precise
(Figure 2). As only 12.2% and 5.62% of polymyxin B was

RESULTS

Demographic data of the 24 intensive care patients are present-

ed in Table 1. TBWs ranged from 41 kg to 110 kg in 23 pa-
tients, and 1 patient was extremely obese at 250 kg. A large
range of renal functions was observed. The physician-selected
dose of polymyxin B was 0.45–3.38 mg/kg/day; 23 patients re-
ceived the drug every 12 hours, whereas the patient who was
prescribed the lowest daily dose received it every 24 hours.
Plasma polymyxin B concentration-time profiles arising
from the physician-selected dosage regimens are presented in
Figure 1. The AUC over a day (AUC0–24 hours) from the popula-
tion PK analysis was 66.9 ± 21.6 mg/hour/L (range, 16.4–117
mg/hour/L), and therefore the average steady-state plasma con-
centration (Css,avg, ie, AUC0–24 hours/24 hours) was 2.79 ± 0.90
mg/L (range, 0.68–4.88 mg/L). The polymyxin B concentra- Figure 1. Plasma concentration-time profiles of polymyxin B in 24 pa-
tions from the 2 CVVHD patients [11] were within the range of tients. Concentrations from the 2 patients undergoing continuous venove-
concentrations observed in non-CVVHD patients (Figure 1). nous hemodialysis [11] are shown by filled symbols.

526 • CID 2013:57 (15 August) • Sandri et al

 Downloaded from http://cid.oxfordjournals.org/ by guest on January 9, 2016
Figure 2. Upper panels: Representative fits of the pharmacokinetic model to the plasma polymyxin B concentration-time data of individual patients.
Lower panels: Individual (left) and population (right) fitted versus observed polymyxin B concentrations.

removed by CVVHD [11], the concentration-time profiles from

Table 2. Population Pharmacokinetic Parameter Estimates, all patients (ie, on and not on renal replacement therapy) were
Between-Subject Variability, and Precision of Estimatesa successfully described simultaneously with one set of popula-
tion PK parameter estimates (Table 2). Linear scaling of clear-
Parameter Population Between-Subject Standard Error
ances and volumes of distribution by TBW reduced the
(Unit) Estimate Variability (%CV) (%SE)
unexplained BSV (%CV) by 3.4% for CL and 41.7% for the
CL (L/h/kg)b 0.0276 32.4 7.49
central volume of distribution (V1). Allometric scaling of CL
V1 (L/kg) 0.0939 73.3 23.6
(scaled by TBW0.75) performed slightly better than linear
V2 (L/kg) 0.330 70.1 19.5
CLic (L/h/kg) 0.146 50.4 22.2
scaling. Linear and allometric scaling by LBW performed simi-
SDintercept (mg/L) 0.0392 larly well as TBW. For simpler clinical implementation, the
SDslope 9.59% results from the model with linear scaling by TBW are reported
in Table 2. The BSV in clearance (CV 32.4%, Table 2) was re-
Abbreviations: CL, total body clearance; CLic, intercompartmental clearance;
CV, coefficient of variation; SDintercept, additive residual error; SDslope, markably low in this critically ill patient population with a wide
proportional residual error; SE, standard error; V1, central volume of distribu- range of body weights and renal functions. After scaling of
tion; V2, peripheral volume of distribution.
a clearances and volumes of distribution by TBW, the parameter
Unscaled population estimates: CL = 1.87 L/h, V1 = 6.35 L, V2 = 22.3 L,
CLic = 9.86 L/h, t1/2β = 11.9 h. estimates for the 2 patients receiving CVVHD, including the
b
For the 2 continuous venovenous hemodialysis (CVVHD) patients, the patient with 250-kg TBW, were within the range of estimates
estimated total body clearance includes the CVVHD clearance. The CVVHD
clearances in the 2 renal replacement therapy patients were only 0.26 L/h and
from the other patients, as shown for CL in Figure 3. Important-
0.37 L/h [11]. ly, neither the unscaled (P = .68, r 2 = 0.008) nor scaled (P = .22,

Pharmacokinetics of Polymyxin B • CID 2013:57 (15 August) • 527

Figure 3. Individual polymyxin B clearance estimates versus creatinine clearance. Polymyxin B clearance was either unscaled (L/hour, left panel) or
scaled by total body weight (L/hour/kg, right panel). Open circles represent patients not on hemodialysis, the filled diamond represents the continuous ve-
novenous hemodialysis (CVVHD) patient who weighed 250 kg, and the filled triangle the lean CVVHD patient.

Downloaded from http://cid.oxfordjournals.org/ by guest on January 9, 2016

r 2 = 0.068) polymyxin B CL showed any relationship with CrCL 4.04% (range, 0.98%–17.4%), and the median renal clearance
(Figure 3). Furthermore, no relationships were identified between (CLR) was 0.061 L/hour (range, 0.018–0.377 L/hour). The CLR
CL of polymyxin B and APACHE II score, sex, age, or serum of polymyxin B was a small percentage of CLGF in every patient
albumin concentration. (median, 9.7%; range, 0.75%–27.9%). Thus, a large percentage
The median unbound fraction in plasma of polymyxin B was of polymyxin B filtered at the glomerulus was reabsorbed
0.42 (range, 0.26–0.64; n = 23) and was independent of concen- (median, 90.3%; range, 72.1%–99.2%); by comparison, the
tration (P > .05, r 2 = 0.13). The mean (±SD) of the AUC for median percentage of filtered water that was reabsorbed was
unbound polymyxin B (fAUC0–24 hours) was 29.2 ± 12.0 mg 96.4% (range, 85.7%–99.0%). There was a trend for the percent-
hour/L (range, 6.05–60.5 mg hour/L). Urinary excretion data age of filtered polymyxin B that was reabsorbed to increase
were available from 17 patients. The median percentage of the with creatinine clearance; the result being that there was a
polymyxin B dose that was excreted unchanged in urine was strong linear relationship (P < .0001, r 2 = 0.90) between the
amount of polymyxin B reabsorbed per day, scaled to the daily
dose, to increase with CrCL (Figure 4). The exposure to poly-
myxin B with various dosage regimens, predicted from the
Monte Carlo simulations, is quantified in Table 3.

DISCUSSION

This polymyxin B population PK study has made a significant

contribution to understanding of how to optimize the clinical
use of this important last-line antibiotic in critically ill patients.
It is the first study to demonstrate that total body weight is a
patient characteristic that influences polymyxin B PK and that
the total body clearance, and hence daily dose requirement, of
polymyxin B is not affected by kidney function.
Scaling of clearances and volumes of distribution by TBW
resulted in unbiased fits for all patients, including the extremely
obese patient (250 kg), and an overall reduction in the BSV.
Figure 4. Relationship between the percentage of the total daily dose Thus, loading and maintenance doses for polymyxin B are best
of polymyxin B undergoing renal tubular reabsorption and creatinine clear- scaled by TBW; this finding informed the manner in which the
ance. Monte Carlo simulations were conducted.

528 • CID 2013:57 (15 August) • Sandri et al

Table 3. Polymyxin B Exposure for 6 Different Dosage Regimens on the First and Fourth Day of Treatment Based on Monte Carlo Simula-
tionsa

Cmax (mg/L)b Cmin (mg/L)b AUC0–24 hours (mg·h/L)

Day P10 P50 P90 P10 P50 P90 P10 P50 P90
1.25 mg/kg q12h as 1-h infusion
Day 1 2.59 5.17 9.38 0.79 0.903 1.48 25.0 46.4 81.1
Day 4 4.34 7.09 11.3 1.06 1.87 3.08 44.3 72.0 114
2 mg/kg loading as 2-h infusion, followed by 1.25 mg/kg q12h as 1-h infusionc
Day 1 3.06 5.71 10.5 0.86 1.48 2.43 34.0 61.7 108
Day 4 4.35 7.06 11.3 1.07 1.90 3.11 44.7 72.7 115
1.5 mg/kg q12h as 1-h infusion
Day 1 3.11 6.21 11.25 0.620 1.08 1.77 29.9 55.7 97.3
Day 4 5.20 8.51 13.56 1.27 2.25 3.69 53.1 86.4 137.3
2.5 mg/kg loading as 2-h infusion followed by 1.5 mg/kg q12h as 1-h infusionc
Day 1 3.95 7.39 13.5 1.11 1.92 3.15 43.4 78.9 137.9
Day 4 5.40 8.76 14.0 1.33 2.36 3.87 55.5 90.4 142.7
2.5 mg/kg/d as continuous infusion
Day 1 ... ... ... ... ... ... 20.4 36.9 63.4

Downloaded from http://cid.oxfordjournals.org/ by guest on January 9, 2016

Day 4 ... ... ... ... ... ... 47.0 72.3 110
2 mg/kg loading as 2-h infusion, then 2.5 mg/kg/d as continuous infusiond
Day 1 ... ... ... ... ... ... 42.4 75.8 128
Day 4 ... ... ... ... ... ... 48.0 73.9 111

Abbreviations: AUC0–24 hours, area under the plasma concentration-time curve over 24 hours; Cmax, maximum polymyxin B concentration; Cmin, minimum polymyxin
B concentration; P10, 10th percentile; P50, 50th percentile; P90, 90th percentile; q12h, every 12 hours.
a
All values refer to total polymyxin B concentration.
b
Cmax and Cmin on day 1 and day 4 refer to dose 1 and dose 8. The Cmin on day 1 is the concentration at the end of the first dosage interval.
c
First maintenance dose administered 12 hours after the loading dose.
d
Continuous infusion commenced immediately after the loading dose.

Polymyxin B CL scaled by TBW displayed only modest in- doses of polymyxin B, particularly in patients with declining
terindividual variability, particularly given the very diverse de- renal function. It should be noted that the benefit of higher
mographics including sex, age, renal function, and severity of polymyxin B dosage regimens (≥200 mg/day) on overall hospi-
illness. Notably, despite the wide range of renal function across tal mortality remained even for patients who developed moder-
the patients, this patient characteristic was not a determinant of ate or severe renal impairment during therapy [19].
polymyxin B CL (Figure 3). That renal function did not influ- The Monte Carlo simulations indicated that dosage regimens
ence polymyxin B CL is in keeping with the fact that only a not involving a loading dose resulted in exposure to polymyxin
small percentage of the dose was excreted in urine as un- B across day 1 that was substantially lower than the exposure
changed drug, as we have previously reported [8]. Non–renal achieved on day 4 (ie, at steady state). It should be noted that
clearance has also been demonstrated to be the major elimina- even though loading doses have been proposed for CMS,
tion pathway of polymyxins in rats [16, 18]. several hours’ delay occurs in the achievement of Cmax of the
As polymyxin B CL was not related to CrCL (Figure 3), its daily antibacterial entity colistin because of the slow formation from
doses should not be based on renal function. This contrasts the prodrug CMS [6, 7, 20, 21]. Our data clearly show the po-
strongly with colistin wherein daily doses need to be tailored to tential PK/PD advantage of polymyxin B versus CMS after in-
renal function, because the latter is administered as the predomi- travenous administration and the importance of employing a
nantly renally eliminated prodrug CMS [7]. Decreasing daily doses loading dose to achieve optimal plasma exposure of polymyxin
of polymyxin B for patients with poor renal function may lead to B as soon as possible.
suboptimal plasma exposure, with potentially adverse consequenc- The fAUC/minimum inhibitory concentration (MIC) has
es on clinical and microbiological outcomes and development of been shown to be the most predictive PK/PD index for the in
resistance. Physicians should balance the risk of polymyxin- vivo antibacterial activity of colistin [22, 23]. In the thigh infec-
induced nephrotoxicity against the benefit of maintaining adequate tion model, the fAUC/MIC values for 2-log bacterial killing

Pharmacokinetics of Polymyxin B • CID 2013:57 (15 August) • 529

were approximately 20 for Pseudomonas aeruginosa and Acine- tubular reabsorption may serve to decrease the potential for
tobacter baumannii. Therefore, assuming that these PK/PD nephrotoxicity.
data for colistin are similar for polymyxin B, our Monte Carlo A potential limitation of our study was the use of CrCL esti-
simulations using an fu of 0.42 show that 1.5 mg/kg/12 hours (ie, mated by the Cockcroft-Gault equation, which has not been
3 mg/kg/day) would reach an fAUC/MIC of approximately 20 validated in critically ill patients. Nonetheless, this limitation
on day 4 in approximately 50% of patients when the causative does not affect our results because there was clearly no relation
pathogen MIC is 2 mg/L (Table 3). Thus, for severe infections between renal function and polymyxin B CL. In addition, only
caused by organisms with polymyxin B MIC of ≤2 mg/L, regi- 2 patients on renal replacement therapy were included in the
mens with a “high” daily dose (eg, 3 mg/kg/day), with a loading population PK analysis.
dose, should be considered. Nonetheless, it is very likely that the In conclusion, this is the first population PK study demon-
currently recommended dosage regimens (up to 2.5 mg/kg/day) strating that doses of intravenous polymyxin B are best scaled
are appropriate for less severe infections, or when the polymyxin by total body weight and should not be based upon renal func-
B MIC of the pathogen is ≤1 mg/L. However, for pathogens with tion. For patients on renal replacement therapy, dosage adjust-
MICs of 4 mg/L, only a very small proportion of patients will ments are not recommended at this time. Further clinical
reach an fAUC/MIC of approximately 20, even with 3 mg/kg/day. studies on polymyxin B PK/PD are urgently needed.
Since >3 mg/kg/day cannot be recommended at this time due to
the lack of clinical data on safety, combination therapy should be
Notes
considered for severe infections caused by such pathogens.

Downloaded from http://cid.oxfordjournals.org/ by guest on January 9, 2016

A retrospective cohort study showed that ≥200 mg/day poly- Financial support. This study was supported by Fundo de Incentivo a
Pesquisa e Eventos do Hospital de Clínicas de Porto Alegre (08–686) and
myxin B was independently associated with lower hospital
the National Council for Scientific and Technological Development
mortality [19]. Because 200 mg/day corresponds to 2.5, 2.85, (CNPq), Ministry of Science and Technology, Brazil (474740/2008-0).
and 3.0 mg/kg per day in patients weighing 80, 70, and 65 kg, A. P. Z. is a research fellow from the CNPq, Ministry of Science and Tech-
respectively, ≥200 mg/day is very likely in accordance with the nology, Brazil (305263/2011–0). J. L. is an Australian National Health and
Medical Research Council (NHMRC) Senior Research Fellow, and this
dosage regimens associated with bactericidal activity of polymyx- study was partially supported by NHMRC project 1026109.
ins, according to the data from mouse infection models [22, 23] Potential conflicts of interest. A. P. Z. has received consultancy fees
and our Monte Carlo simulations (Table 3). from Pfizer, Eurofarma, and Forest Laboratories. D. R. F. has received con-
sultancy fees from Pfizer. All other authors report no potential conflicts.
Although renal clearance contributes in a minor way to poly- All authors have submitted the ICMJE Form for Disclosure of Potential
myxin B CL, the processing of the drug within the kidney is ex- Conflicts of Interest. Conflicts that the editors consider relevant to the
tremely important as nephrotoxicity is the major dose-limiting content of the manuscript have been disclosed.
adverse effect [19, 24]. In the present study, urinary excretion
and plasma protein binding data were available for 16 patients References
with a wide range of renal functions. In all of these patients,
1. Li J, Nation RL, Turnidge JD, et al. Colistin: the re-emerging antibiotic
polymyxin B CLR was only a very small percentage of the antic- for multidrug-resistant gram-negative bacterial infections. Lancet
ipated filtration clearance of the drug at the glomeruli, indicat- Infect Dis 2006; 6:589–601.
ing that polymyxin B was subject to extensive net tubular 2. Zavascki AP, Goldani LZ, Li J, Nation RL. Polymyxin B for the treat-
ment of multidrug-resistant pathogens: a critical review. J Antimicrob
reabsorption, as has been reported previously for 4 other pa-
Chemother 2007; 60:1206–15.
tients [8] and for colistin in rats [16]. The trend for the percent- 3. Falagas ME, Kasiakou SK. Colistin: the revival of polymyxins for the
age of filtered polymyxin B reabsorbed to increase with CrCL is management of multidrug-resistant gram-negative bacterial infections.
consistent with the intact nephron hypothesis [25]. The rela- Clin Infect Dis 2005; 40:1333–41.
4. Landman D, Georgescu C, Martin DA, Quale J. Polymyxins revisited.
tionship between the percentage of the daily dose that was reab- Clin Microbiol Rev 2008; 21:449–65.
sorbed and CrCL (Figure 4) highlights the degree of exposure 5. Bergen PJ, Li J, Rayner CR, Nation RL. Colistin methanesulfonate is an
of tubular cells to polymyxin B. For example, for a patient with inactive prodrug of colistin against Pseudomonas aeruginosa. Antimi-
crob Agents Chemother 2006; 50:1953–8.
a CrCL ≥100 mL/min, remarkably, an amount equivalent to 6. Mohamed AF, Karaiskos I, Plachouras D, et al. Application of a loading
the daily dose or more was reabsorbed through the tubular cells dose of colistin methanesulfonate in critically ill patients: population
per day. In the case of a patient with a lower CrCL, the amount pharmacokinetics, protein binding, and prediction of bacterial kill.
Antimicrob Agents Chemother 2012; 56:4241–9.
reabsorbed per day can still be a substantial percentage of the
7. Garonzik SM, Li J, Thamlikitkul V, et al. Population pharmacokinetics
daily dose, especially given the likely fewer number of function- of colistin methanesulfonate and formed colistin in critically ill patients
ing nephrons in such a patient [25]. Thus, there is substantial from a multicenter study provide dosing suggestions for various catego-
recycling of polymyxin B between tubular urine and the systemic ries of patients. Antimicrob Agents Chemother 2011; 55:3284–94.
8. Zavascki AP, Goldani LZ, Cao G, et al. Pharmacokinetics of intrave-
circulation, which results in extensive intracellular exposure to nous polymyxin B in critically ill patients. Clin Infect Dis 2008;
polymyxin B. Our data (Figure 4) suggest that inhibition of 47:1298–304.

530 • CID 2013:57 (15 August) • Sandri et al

 9. Kwa AL, Lim TP, Low JG, et al. Pharmacokinetics of polymyxin B1 in 18. Abdelraouf K, Braggs KH, Yin T, Truong LD, Hu M, Tam VH. Charac-
patients with multidrug-resistant gram-negative bacterial infections. terization of polymyxin B-induced nephrotoxicity: implications for
Diagn Microbiol Infect Dis 2008; 60:163–7. dosing regimen design. Antimicrob Agents Chemother 2012; 56:4625–9.
10. Kwa AL, Abdelraouf K, Low JG, Tam VH. Pharmacokinetics of poly- 19. Elias LS, Konzen D, Krebs JM, Zavascki AP. The impact of polymyxin
myxin B in a patient with renal insufficiency: a case report. Clin Infect B dosage on in-hospital mortality of patients treated with this antibiot-
Dis 2011; 52:1280–1. ic. J Antimicrob Chemother 2010; 65:2231–7.
11. Sandri AM, Landersdorfer CB, Jacob J, et al. Pharmacokinetics of poly- 20. Plachouras D, Karvanen M, Friberg LE, et al. Population pharmacoki-
myxin B in patients on continuous venovenous haemodialysis. J Anti- netic analysis of colistin methanesulfonate and colistin after intrave-
microb Chemother 2013; 68:674–7. nous administration in critically ill patients with infections caused by
12. Bauer RJ. S-ADAPT/MCPEM user’s guide (version 1.57). Software for gram-negative bacteria. Antimicrob Agents Chemother 2009; 53:
pharmacokinetic, pharmacodynamic MCPEM and population data 3430–6.
analysis. Berkeley, CA, 2011. 21. Markou N, Fousteri M, Markantonis SL, et al. Colistin pharmacokinet-
13. Bulitta JB, Bingolbali A, Shin BS, Landersdorfer CB. Development of a ics in intensive care unit patients on continuous venovenous haemodia-
new pre- and post-processing tool (SADAPT-TRAN) for nonlinear filtration: an observational study. J Antimicrob Chemother 2012; 67:
mixed-effects modeling in S-ADAPT. AAPS J 2011; 13:201–11. 2459–62.
14. Bulitta JB, Landersdorfer CB. Performance and robustness of the 22. Dudhani RV, Turnidge JD, Coulthard K, et al. Elucidation of the phar-
Monte Carlo importance sampling algorithm using parallelized macokinetic/pharmacodynamic determinant of colistin activity against
S-ADAPT for basic and complex mechanistic models. AAPS J 2011; Pseudomonas aeruginosa in murine thigh and lung infection models.
13:212–26. Antimicrob Agents Chemother 2010; 54:1117–24.
15. Janmahasatian S, Duffull SB, Ash S, Ward LC, Byrne NM, Green B. 23. Dudhani RV, Turnidge JD, Nation RL, Li J. fAUC/MIC is the most pre-
Quantification of lean bodyweight. Clin Pharmacokinet 2005; 44: dictive pharmacokinetic/pharmacodynamic index of colistin against
1051–65. Acinetobacter baumannii in murine thigh and lung infection models.
16. Li J, Milne RW, Nation RL, Turnidge JD, Smeaton TC, Coulthard K. J Antimicrob Chemother 2010; 65:1984–90.
Use of high-performance liquid chromatography to study the pharma- 24. Kubin CJ, Ellman TM, Phadke V, Haynes LJ, Calfee DP, Yin MT. Inci-

Downloaded from http://cid.oxfordjournals.org/ by guest on January 9, 2016

cokinetics of colistin sulfate in rats following intravenous administra- dence and predictors of acute kidney injury associated with intravenous
tion. Antimicrob Agents Chemother 2003; 47:1766–70. polymyxin B therapy. J Infect 2012; 65:80–7.
17. Beal SL, Sheiner LB, Boeckmann AJ, Bauer RJ. NONMEM user’s guides 25. Bricker NS. On the meaning of the intact nephron hypothesis. Am J
(1989–2009). Ellicott City, MD: Icon Development Solutions, 2009. Med 1969; 46:1–11.

Pharmacokinetics of Polymyxin B • CID 2013:57 (15 August) • 531