J Antimicrob Chemother 2014; 69: 180 189

doi:10.1093/jac/dkt304 Advance Access publication 1 August 2013

Population pharmacokinetics of piperacillin and tazobactam in critically

ill patients undergoing continuous renal replacement therapy:
application to pharmacokinetic/pharmacodynamic analysis
Eduardo Asn-Prieto1,2, Alicia Rodrguez-Gascon1,2, Inaki F. Troconiz3, Amaia Soraluce1,2, Javier Maynar4,
Jose Angel Sanchez-Izquierdo5 and Arantxazu Isla1,2*
1
Pharmacokinetics, Nanotechnology and Gene Therapy Group, Faculty of Pharmacy, University of the Basque Country UPV/EHU,
Vitoria-Gasteiz, Spain; 2Centro de Investigacion Lascaray ikergunea, University of the Basque Country UPV/EHU, Vitoria-Gasteiz, Spain;
3
Department of Pharmacy and Pharmaceutical Technology, Faculty of Pharmacy, University of Navarra, Pamplona, Spain; 4Intensive Care
Unit, University Hospital of Alava (sede Santiago), Vitoria-Gasteiz, Spain; 5Intensive Care Unit, Doce de Octubre Hospital, Madrid, Spain

Received 1 March 2013; returned 21 April 2013; revised 30 May 2013; accepted 1 June 2013
Objectives: To evaluate the pharmacokinetics of piperacillin/tazobactam in critically ill patients undergoing continuous renal replacement therapy (CRRT) and to assess the success of the therapy against susceptible bacteria.
Patients and methods: Sixteen patients undergoing CRRT with different degrees of renal function were included in
the study. Blood and ultrafiltrate samples were drawn after administration of piperacillin/tazobactam (4/0.5 g)
every 4, 6 or 8 h. The data were analysed by a population approach using NONMEM 7.2. The probability of target
attainment (PTA) of maintaining free piperacillin levels above the MIC during the entire dosing interval was estimated by simulation of intermittent and continuous infusions.
Results: The pharmacokinetics of piperacillin and tazobactam were best described by two-compartment models
where the elimination of both drugs was conditioned by renal [dependent on creatinine clearance (CLCR)], nonrenal and extracorporeal clearances. A 20 min infusion of piperacillin/tazobactam administered every 6 h provided
high PTAs against MICs 32 mg/L in patients with severe renal failure. In patients with normal or moderate renal
function PTAs 90% were only obtained up to MICs 8 mg/L with short infusions. However, simulating continuous
infusion, higher probabilities of success were obtained against MICs of 32 and 16 mg/L when CLCR was 50 and
100 mL/min, respectively.
Conclusions: Population pharmacokinetic models have been developed and validated for piperacillin and tazobactam. Based on the pharmacokinetic/pharmacodynamic analysis, dosing recommendations are given considering
the residual renal function of the patient and the MIC for the isolated bacteria.
Keywords: antimicrobial therapy, b-lactams, PD, PK, population pharmacokinetics, renal replacement therapy, intensive care

Introduction
Piperacillin, administered concomitantly with tazobactam (a
b-lactamase inhibitor), is one of the most frequently used intravenous antimicrobials for the treatment of infections in critically ill
patients.1 It shows broad-spectrum antibacterial activity against
Gram-positive and Gram-negative aerobic and anaerobic bacteria,
including Pseudomonas aeruginosa.2
The population pharmacokinetics of piperacillin/tazobactam
have been studied previously,1,3 but not in critically ill patients
with different degrees of renal function being treated with continuous renal replacement therapy (CRRT). These procedures involve

pumping blood from the patients circulation through an extracorporeal circuit incorporating a synthetic membrane. Plasma water
and solutes with small molecular weight are removed from the
blood. Both piperacillin and tazobactam have small molecular
weight and low protein binding, and are significantly removed by
CRRT.2,4 This fact should be taken into account for population pharmacokinetic model construction, which should evaluate the influence of extracorporeal clearance on the total body clearance of
drugs.
Considering the reported variability in piperacillin and tazobactam pharmacokinetic parameters during critical illness and the

# The Author 2013. Published by Oxford University Press on behalf of the British Society for Antimicrobial Chemotherapy. All rights reserved.
For Permissions, please e-mail: journals.permissions@oup.com

180

Downloaded from http://jac.oxfordjournals.org/ by guest on December 6, 2015

*Corresponding author. Laboratory of Pharmacy and Pharmaceutical Technology, Faculty of Pharmacy, University of the Basque Country UPV/EHU,
Paseo de la Universidad 7, 01006 Vitoria-Gasteiz, Spain. Tel: +945-01-34-69; Fax: +945-01-30-40; E-mail: arantxa.isla@ehu.es

JAC

Piperacillin/tazobactam population PK/PD during CRRT

lack of population pharmacokinetic information of these drugs

during CRRT, the first objective of this work was to provide a
better understanding of piperacillin/tazobactam disposition in
this clinical situation and to assess factors that significantly influence drug pharmacokinetics that affect drug exposure. A second
objective was to elucidate whether the standard dose of 4/0.5 g
piperacillin/tazobactam administered intravenously with different
dosing intervals is able to maximize the opportunity for therapeutic
outcomes in different groups of patients, by determining the probability of maintaining free drug concentrations above different MIC
values (fT.MIC).

during continuous renal replacement therapy, was calculated as the ratio

of the area under the ultrafiltrate concentration curve to the area under
the serum concentration curve,6 obtained by no-compartmental fitting.
This coefficient can also be calculated as the ratio of the ultrafiltrate concentration to serum concentration at each timepoint.6 Extracorporeal clearance
(CLEC) for piperacillin was set as Sc multiplied by ultrafiltrate flow (Quf; also
called effluent rate). In some cases, the unbound fraction (fu) may be used
instead of Sc when correlation between them has been found and both
parameters show similar values,6 as occurred with tazobactam in the
present study.

Patients and methods

Table 2 shows the MIC values of piperacillin/tazobactam for the 21 strains

isolated from patients. MICs were determined in accordance with the
CLSI (formerly NCCLS) guidelines (M7-A5). Broth microdilution tests were
performed in custom-dried 96-well microdilution trays (Sensititre Division,
Accumed International, Westlake, OH, USA).

Study design and settings

Renal replacement therapy

All patients were undergoing continuous venovenous haemofiltration. Vascular access was obtained with 13.5 FG dual lumen catheters (Niagara,
Bard, Ontario, Canada). A Prisma (Hospal, Lyon, France) machine was
used with a 0.9 m2 AN69 acrylonitrile and sodium methyl sulfonate copolymer filter (Prisma M100 Hospal, Lyon, France). Blood flow was kept between
140 and 230 mL/min and ultrafiltrate flow was maintained between 1000
and 2150 mL/h (Table 1). The ultrafiltrate was replaced as clinically indicated with Ringers lactate at rates ranging from 900 to 2100 mL/h.
Replacement fluid was delivered pre-filter.

Table 1. Summary of patient characteristics

Patient characteristic
Continuous covariates
age (years)
body weight (kg)
Quf (mL/h)
Qrep (mL/h)
Qb (mL/min)
CLCR (mL/min)
CLCR in septic patients
CLCR in polytraumatized patients
fu_piperacillin (%)
fu_tazobactam (%)
serum total protein concentration (g/dL)
serum albumin concentration (g/dL)
serum bilirubin concentration (mg/L)
Sc piperacillina
Sc tazobactama
Categorical covariates
gender (male/female)
SOFA score
APACHE II score
type of patient (septic/polytraumatized)

Mean (SD)

Range

57 (16)
74 (8)
1619 (425)
1538 (433)
172 (33)
43 (34)
30 (23)
73 (38)
78 (13)
66 (22)
4.7 (0.9)
2.15 (0.47)
2.13 (1.89)
0.37 (0.25)
0.76 (0.26)

18 77
60 90
1000 2150
900 2100
140 230
1.3 110
1.3 77.6
10 110
48 95
16 97
3.6 6.5
1.5 3.1
0.4 6.7
0.11 1
0.39 1.27

12/4
11 (3)
22 (6)
11/5

8 17
10 36

Qrep, repository flow; Qb, blood flow to the filter; SOFA, Sepsis-related Organ
Failure Assessment;29 APACHE II, Acute Physiology and Chronic Health
Evaluation II.30
a
Estimated as the ratio of area under the ultrafiltrate concentration curve to
area under the serum concentration curve.

Drug administration, sampling procedure and analytical

method
All patients were given piperacillin/tazobactam (4/0.5 g) in a 20 min intravenous infusion every 4, 6 or 8 h. A mean of 11 previous doses were
Table 2. Bacteria isolated from patients and their susceptibilities,
expressed as MICs
Bacteria
Acinetobacter baumannii
Aeromonas hydrophila
Bacteroides vulgatus
Burkholderia cepacia
Citrobacter diversus
Enterococcus faecalis
Escherichia coli
Escherichia coli
Hafnia alvei
Klebsiella pneumoniae
Pseudomonas aeruginosa
Pseudomonas aeruginosa
Serratia marcescens
Staphylococcus epidermidis,
coagulase-negative
Staphylococcus aureus
Stenotrophomonas maltophilia

Number of isolates

MIC (mg/L)

1
1
1
1
1
2
1
3
1
1
3
1
1
1

32
2
16
64
8
16
0.25
8
2
8
16
64
64
4

1
1

64
ND

ND, not determined.

181

Downloaded from http://jac.oxfordjournals.org/ by guest on December 6, 2015

An open-label prospective study was performed in the intensive care unit

(ICU) at two different hospitals, University Hospital of Alavasede Santiago(Vitoria, Spain) and Doce de Octubre Hospital (Madrid, Spain). All
study procedures were approved by the ethics committees of the participant hospitals and were conducted in accordance with Good Clinical Practice. Written informed consent was required from all patients. A total of 16
patients aged .18 years and needing to be treated with continuous venovenous haemofiltration for at least 20 h were included. Table 1 displays the
demographic data of the patients together with other biochemical data
and health scores. Creatinine clearance (CLCR) was estimated for each
patient with the CockcroftGault equation.5 The sieving coefficient (Sc),
defined as the fraction of the drug eliminated across the membrane

Bacterial isolates and MICs

Asn-Prieto et al.

administered to achieve steady-state concentrations of the drugs before

starting sample collection. Blood and ultrafiltrate samples were drawn at
0, 0.3, 0.5, 0.75, 1, 3, 4, 6 and 8 h after dosing. The sampling procedure
was slightly modified for the patients receiving piperacillin every 4 h. All collected samples were analysed using a validated HPLC UV technique.7
Linearity was established over the concentration range from 5 to
500 mg/L for piperacillin and from 3 to 50 mg/L for tazobactam. The accuracy and precision of intra-day and inter-day (studied over 3 days) assays
were determined at the limit of quantification and at three concentrations
in the range of expected concentration. Calculated concentration never
deviated more than 7% from nominal concentration. The intra-day and
inter-day precision, expressed as coefficient of variation, was always
below 7%. Protein binding was measured at peak and trough plasma
levels in each patient by ultrafiltration using Sartorius Centrisart I filters
(cut-off 10 000) (Sartorius AG, Goettingen, Germany).

Base model
Piperacillin and tazobactam plasma concentrations were modelled separately using NONMEM 7.2 and the FOCE method with INTERACTION. The disposition of the total drug plasma concentration was modelled using
compartmental models. Based on the distributions of the residuals, data
from the two drugs were logarithmically transformed. The selection of
the model was based on the decrease in objective function value (OFV),
the relative standard errors of the parameters, calculated as the ratio
between the standard error and final parameter estimate, and the
goodness-of-fit (GOF) plots. A difference of 3.84, 6.63 and 10.83 points in
the OFV between two nested models differing in one parameter was
significant at P values of 0.05, 0.01 and 0.001, respectively. Inter-individual
variability (IIV) was modelled exponentially and residual error with an
additive model on a logarithmic scale. At this step, the significance of the
off-diagonal elements of the V variance covariance matrix were also
explored.

Covariate selection
Once a base model was selected, patient characteristics were explored for
influence on pharmacokinetic parameters. Table 1 displays the covariates
studied for each drug. Continuous covariates were centred at the median
value of the population. Depending on the method used for Sc estimation,
extracorporeal clearance (CLEC) was studied as a fixed value for each patient
or as a time-varying covariate, following the methodology described by
Wahlby et al.8 The effect of pre-dilution by the replacement fluid on CLEC
was also studied.6 The identification of covariate candidates was carried
out by stepwise covariate model building (SCM tool in PsN 3.5.3). The following levels of significance were used during the forward inclusion, backward
deletion and revaluation of IIV, respectively: P,0.05, P, 0.01 and
P,0.001. Simulations and GOF plots were also used to support the inclusion
and deletion steps. Finally, piperacillin and tazobactam final models were
modelled together to study the possibility of interaction between the two
drugs.

Model evaluation
For model evaluation, diagnostic plots or GOF plots were the first indicator of
suitability. The population and individual predictions versus observations,
the conditioned weighted residual errors versus time and the normalized
prediction distribution errors versus population predictions for each drug
were the GOF plots constructed. Parameter precision was evaluated by performing a 2000 dataset bootstrap (Bootstrap tool in PsN 3.5.3).9 A

182

Pharmacokinetic/pharmacodynamic analysis
Different dosing regimens were evaluated in order to predict the probability
of reaching the pharmacokinetic/pharmacodynamic target. Since the
patients included in the study were critically ill, a pharmacokinetic/pharmacodynamic target of fT.MIC of 100% was considered.12,13 Using the same
dosing regimens given to patients (4/0.5 g of piperacillin/tazobactam in a
20 min infusion every 4, 6 or 8 h), 1000 studies of 1000 subjects with different CLCR values (10, 50 or 100 mL/min) were simulated with the piperacillin
final model. Moreover, extended (4 h) infusions of 4/0.5 g piperacillin/tazobactam every 6 or 8 h and continuous infusions containing 12,16 and
24g/day were also evaluated for patients with CLCR values .50 mL/min.
The target was attained when the simulated Cmin of piperacillin in plasma
at steady state was greater than the MIC. The probability of target attainment (PTA) was calculated over a range of doubling MICs between 0.125
and 128 mg/L. For this estimation, plasma fu for piperacillin, Quf and Sc
were introduced as mean and standard deviation following the distribution
of the original dataset for the corresponding subpopulation.
Pharmacokinetic/pharmacodynamic evaluation was only carried out
for piperacillin since it has antimicrobial activity, whereas tazobactam is
administered as an inhibitor of b-lactamases.

Results
Sixteen critically ill patients with a mean CLCR of 43 mL/min (range
1.3 110 mL/min) were enrolled in the study. Seven patients
received 4/0.5 g of piperacillin/tazobactam every 8 h, seven
patients received 4/0.5 g of piperacillin/tazobactam every 6 h,
one patient received 4/0.5 g of piperacillin/tazobactam every 4 h
and one patient received 4 g of piperacillin every 4 h and 0.5 g of
tazobactam every 8 h. Approximately seven plasma samples per
patient and drug were analysed.

Piperacillin model
Piperacillin plasma concentrations were best described by a twocompartment model. In the base model, IIV was included for CL,
the central compartment volume of distribution (V1) and the peripheral compartment volume of distribution (V2). Covariance
between random effects CL and V1 was found to be significant
(P, 0.05). The piperacillin CL was estimated as the sum of a nonrenal component (CLNR), a renal component (CLR) conditioned by
the CLCR and CLEC. CLEC was studied in different ways. Firstly, it was

Downloaded from http://jac.oxfordjournals.org/ by guest on December 6, 2015

Model estimation and population pharmacokinetic/

pharmacodynamic analysis

prediction- and variability-corrected visual predictive check (pvcVPC) was

used to explore model performance of the selected model (VPC tool in
PsN 3.5.3). By using pvcVPC, both the observations and model predictions
were normalized for the typical model predictions and the expected variability within individuals was also normalized with respect to the typical
model variability.10 pvcVPCs were constructed with the 5th, 50th and
95th percentiles for the observed data (90% inter-percentile range).
Then, 1000 datasets were simulated from the final model parameter estimates, and the 95% CIs for the 5th, 50th and 95th percentiles based on the
simulated datasets were calculated and represented together for visual
inspection (using the xpose4 package in R 2.12.0).
External validation was conducted by assessing the ability of the population model to predict concentrations and clearances from previously published data.11 Considering the characteristics of the subjects included in the
cited study, 1000 datasets, each containing 250 subjects were simulated.
The 5th, 50th and 95th percentiles of total body clearance (CL), free
maximum concentration (Cmax) and free minimum concentration (Cmin)
were calculated for the simulated data and were compared with the
values provided by the authors.

JAC

Piperacillin/tazobactam population PK/PD during CRRT

Tazobactam model
A two-compartment model was also the best to describe the
observed data for tazobactam. IIV was included in CL and V1 and
no correlation was detected. Tazobactam CL was coded as for
piperacillin. The CL of tazobactam was the result of the sum of
CLNR, CLR (affected by CLCR) and CLEC. Due to the lack of ultrafiltrate
data, Sc could not be estimated in one patient. The observed Sc for
tazobactam, unlike that for piperacillin, was similar to the unbound
fraction.2 Therefore, fu was used instead of Sc for the calculation of
CLEC (CLEC fuQuf ). The type of patient turned out to be a statistically significant covariate for V1, with this parameter showing higher
values in polytraumatized patients (typical value V1 51.93 L)
than in septic ones (V1 17.45 L). No other covariate resulted in a
relevant reduction in the OFV. The IIV was diminished from 64%
to 25% in the case of CL and from 62% to 36% for V1. Table 3
also displays the population pharmacokinetic parameter estimates for tazobactam.
No interaction between piperacillin and tazobactam was
detected when they were modelled simultaneously.

Table 3. Population pharmacokinetic model estimates, shrinkage valuesa

and bootstrap results for piperacillin and tazobactam after intravenous
infusion

Parameter
Piperacillin
CL (L/h)CLNR +CLR +CLEC
CLNR
CLR (CLCR/1.95)u
CLEC QufSc
V1 (L)
Q (L/h)
V2 (L) uWT/75.3
IIV_CL (%)
IIV_V1 (%)
correlation of IIV_CL with IIV_V1
(%)
residual error_additive (log
scale)
Tazobactam
CL (L/h)CLNR +CLR +CLEC
CLNR
CLR (CLCR/1.95)u
CLEC Quffu
V1 (L)
V1 septic
V1 polytraumatized
Q (L/h)
V2 (L)
IIV_CL (%)
IIV_V1 (%)
residual error_additive (log
scale)

Estimate, RSE
(%)

Bootstrap, median
(5th95th
percentile)

2.77 (36)
4.55 (22)
b
20.86 (26)
20.53 (30)
21.41 (18)
44 (31)
75 (45)
80 (37)

2.79 (1.144.50)
4.53 (3.076.36)

20.73 (11.6329.55)
20.29 (12.1739.02)
22.45 (16.6131.06)
41 (28 53)
72 (48 107)
82 (20.9138.6)

0.19 (11)

0.19 (0.160.22)

1.33 (38)
4.60 (15)
b

1.37 (0.322.13)
4.63 (3.465.83)

17.45 (10)
51.93 (26)
14.98 (15)
24.18 (16)
25 (28)
36 (54)
0.21 (15)

17.72 (14.8320.98)
51.68 (31.8282.08)
14.90 (10.1619.24)
24.00 (18.6835.68)
23 (15 29)
32 (10 47)
0.21 (0.150.26)

CL, apparent total body clearance of the drug from plasma; Quf , ultrafiltrate
flow; Q, inter-compartment clearance; V2, apparent volume of distribution
of the peripheral compartment; WT, body weight; RSE, relative standard
error; hsh, shrinkage value for a parameter; 1sh, shrinkage value for the
residual error.
a
Piperacillin: CLhsh 22%; V1hsh 3%; 1sh 12%. Tazobactam: CLhsh 1%;
V1hsh 16%; 1sh 11%.
b
CLEC [mean (SD)]: piperacillin 0.64 (0.53) L/h; tazobactam 1.02 (0.35) L/h.

Model evaluation
Diagnostic plots (Figure 1) showed no trend in normalized prediction distribution errors and conditional weighted residual distribution along both time and predicted concentrations for both drugs.
Parameter precision was studied by the bootstrap method and
Table 3 displays the results, showing that all parameters were estimated with good precision. Furthermore, visual inspection of
pvcVPC revealed a good correlation between the percentile intervals obtained by simulations for the final model with the raw
data, as can be seen in Figure 2. When external validation was
carried out, the 50th percentile of the parameter values obtained
by simulation with our model were close to the median values estimated by Bauer et al.,11 as Table 4 shows, confirming the predictive
ability of the model.

Pharmacokinetic/pharmacodynamic analysis
Figure 3 displays the probability of achieving free piperacillin levels
greater than the MIC during the entire dosing interval for the different
short-infusion dosing regimens studied. Large influences of CLCR and
dosing interval on PTA are shown in Figure 3. When 4 g of piperacillin
every 8 h was simulated, PTAs .90% were obtained for MIC values
16, 4 and 1 mg/L in virtual patients with CLCR of 10, 50 and
100 mL/min, respectively. Higher PTAs were obtained when the
dosing interval was reduced. Regarding the administration of 4 g of
piperacillin every 6 h, PTAs .90% were quantified at MIC values
32, 8 and 2 mg/L in virtual patients with CLCR of 10, 50 and
100 mL/min, respectively. Finally, when 4 g of piperacillin every 4 h

183

Downloaded from http://jac.oxfordjournals.org/ by guest on December 6, 2015

studied as a fixed value or as a time-varying covariate, depending

on the methodology used for Sc calculation. A time variation of
this coefficient was not supported by the data and an Sc fixed
value was used per patient. Secondly, blood flow was 7-fold
with respect to replacement fluid and thus the pre-dilution
effect was found to be non-relevant. Therefore, CLEC was
expressed as ScQuf. The full model also supported the inclusion
of body weight in V2 and the type of patient in CLNR and V1, giving
higher values of these parameters for polytraumatized compared
with septic patients. In the backward deletion step, extracting the
type of patient from both CL and V1, an increase smaller than 6.63
was shown (P .0.01). A low increase in the OFV (DOFV, 3.84) was
also shown for body weight in V2, but, as no degrees of freedom
were added to the model and based on simulations it was kept
in the model. Consequently, the selected model included CLEC,
CLCR as a covariate for CL and body weight for V2. The inclusion
of the covariates reduced the IIV in CL from 74% to 44% and in
V2 from 79% to 64%. When re-evaluating the adequacy of variability in the parameters, the IIV for V2 was excluded from the
final model (DOFV was , 10.83 when removing IIV for V2,
P .0.001). Table 3 displays the selected final model and the estimates.

Asn-Prieto et al.

4
3
2

5
npde

(c)
6

4
3

(d)

0
1

2
1

2
1

4
6
Time (h)

4
6
Time (h)

(h)
2

3
npde

(g)

2
CWRES

(f) 4

3
4
5
Population predictions

0
2

0
2

0
1
2
3
4
Individual predictions

1
2
3
Population predictions

1
2
3
Population predictions

Figure 1. GOF plots obtained from the selected final models for piperacillin (top panels) and tazobactam (bottom panels). (a), (b), (e) and (f) correspond to
the scatterplot of the dependent variable versus the individual and population predictions. The dependent variable corresponds to the logarithmic
transformation of the observations. The dotted line is the line of identity and the continuous line corresponds to the linear regression curve. (c) and (g)
correspond to the normalized prediction distribution errors versus the population predictions. (d) and (h) correspond to the conditioned weighted
residuals versus time. The horizontal dotted line is the zero reference line and the continuous line is a smooth non-parametric regression line. npde,
normalized prediction distribution errors; CWRES, conditioned weighted residuals.

was evaluated, the PTA was .90% for MICs 16 and 8 mg/L in
virtual patients with CLCR of 50 and 100 mL/min, respectively.
Figure 4 depicts the PTAs when 4 h and continuous infusions
were simulated for virtual patients with CLCR values of 50 and
100 mL/min. The administration of piperacillin by continuous infusion guarantees PTAs .90% for MICs of 16 mg/L when 12 g/day is
administered in patients with CLCR of 50 mL/min, whereas 16 g/day
is needed for subjects with CLCR of 100 mL/min. In patients with
CLCR of 50 mL/min the daily continuous infusion of 24 g piperacillin
also provides high PTAs against bacteria for which the MICs are
32 mg/L. Extended infusions of 4 g every 6 h provide PTAs .90%
against MICs of 8 and 4 mg/L in virtual patients with CLCR of 50
and 100 mL/min, respectively.

Discussion
Most studies of piperacillin and tazobactam pharmacokinetics in
patients undergoing CRRT include only patients with renal
failure.11,13,14 However, the indications for CRRT go beyond anuric
and renal-impaired patients and include trauma patients, patients
undergoing cardiovascular surgery and septic patients.2 Besides,
due to the heterogeneity of ICU patients it is difficult to predict the
pharmacokinetic profile and, therefore, the outcome of therapy. In
this sense, population modelling could be very useful as it allows
us to quantify the variability shown in these patients and evaluates
the effect of individual physiopathological characteristics in the
pharmacokinetics of the drugs. Therefore, samples from 16 patients
undergoing CRRT with different degrees of renal function were

184

analysed in order to develop population pharmacokinetic models

for both piperacillin and tazobactam, with the purpose of better
understanding the pharmacokinetics of the drugs in this population
and assessing the success of the therapy against susceptible bacteria. To the best of our knowledge, this research is the first that
has studied the population pharmacokinetics and pharmacodynamics of piperacillin/tazobactam combinations in patients
undergoing CRRT with different degrees of renal function.
The pharmacokinetics of piperacillin and tazobactam in the
patients studied is best described by two-compartment models,
as has been reported in other published pharmacokinetic
studies.2,12,14,15 In the selected models, the elimination of piperacillin and tazobactam was conditioned by CLR dependent on CLCR,
CLNR and CLEC. Both drugs are primarily eliminated by the renal
route (.60% of CL),16 but when renal function is affected other
elimination processes gain importance. CLEC had a significant
impact on CL in patients with impaired renal function (.25% of
CL), but elimination of the drugs was not highly affected by CLEC
when renal function remained at normal values (,5%). Recently,
Bauer et al.11 have reported higher values of CL and CLEC than in
this study and they therefore observed a higher impact CLEC
(,50%) due to a higher Quf. However, CLNR estimated by Bauer
et al.11 was similar to CLNR estimated in our study. Therefore, we
studied the influence of the different partial CLs on total CL and
the impact of Quf by simulating 1000 virtual patients based on
our models for both piperacillin and tazobactam (Figure 5). The
impact of CLEC on the total CL of both drugs was highly dependent
on the residual CLCR of the simulated patients, and it was only significant in patients with serious renal damage (CLCR 10 mL/min).

Downloaded from http://jac.oxfordjournals.org/ by guest on December 6, 2015

(e) 4

Ln (observations)

3
4
5
Population predictions

Ln (observations)

2 3 4 5 6
Individual predictions

0
1

2
CWRES

(b)
6

Ln (observations)

Ln (observations)

(a)

Piperacillin/tazobactam population PK/PD during CRRT

Prediction corrected DV

(a)

7
6
5
4
3
2
1

3
4
Time (h)

3
4
Time (h)

Prediction corrected DV

(b) 5
4
3
2
1
0

Figure 2. Results from the pvcVPC for piperacillin (a) and tazobactam (b) up
to 6 h. The dots correspond to the prediction-corrected dependent variable
(DV). The continuous line represents the median and the mid-grey shading
represents the simulation-based 95% CI for the median. Dashed lines
correspond to the observed 5th and 95th percentiles (90%
inter-percentile range) and the simulation-based 95% CIs for the
corresponding percentiles are represented by light grey shading.
Overlapping of mid-grey and light-grey shading results in dark-grey
shading. Patients that were given piperacillin and tazobactam every 4 h
were not included in the pvcVPC.
Table 4. Comparison of the parameter estimates provided in the
literature with the values obtained by simulation with our model
considering the characteristics of the population provided in the cited
study

Bauer et al.,11 median

Simulated dataset, median

(5th 95th percentile)

Piperacillin
CL (L/h)
fCmax (mg/L)
fCmin (mg/L)

3.87
115
54.8

3.67 (1.76 7.68)

128.01 (51.71290.12)
56.15 (22.64135.72)

Tazobactam
CL (L/h)
fCmax (mg/L)
fCmin (mg/L)

2.90
16.3
9.0

2.90 (1.80 4.70)

14.32 (8.50 23.52)
7.89 (4.03 13.98)

Parameter

CL, apparent total body clearance of the drug from plasma; fCmax, peak
concentration of unbound drug in plasma; fCmin, trough concentration of
unbound drug in plasma.

When evaluating the relevance of Quf, it was shown that increases

of 0.5 L/h in this parameter led to increases in CLEC 25%, which
means an improvement of 6% in total CL for both drugs in patients

with CLCR of 10 mL/min. A controversial aspect regarding piperacillin elimination is the possible involvement of a saturable process, as
some authors have suggested,1,17 19 but in our study this could
not be confirmed. Landersdorfer et al. 1 pointed out that the saturation in CLR could be masked by the large variability in patients.
Moreover, these authors suggested that the influence of the saturable elimination of piperacillin is relatively small at a clinically relevant dosage. Another covariate shown to be influential on the
pharmacokinetics of piperacillin is the body weight included in V2.
Besides, the type of patient (septic or polytraumatized) was
found to be significant for the V1 of tazobactam, with higher
values in polytraumatized patients than in septic ones. Actually,
polytraumatized patients received high quantities of intravenous
fluids, which could result in an expanded extracellular compartment.20 The higher variability in protein binding shown for
tazobactam, along with the higher volume of distribution of
tazobactam-free fraction compared with piperacillin described in
previous studies,11 may justify a higher sensitivity of tazobactam
to the expansion of the extracellular compartment than piperacillin. Other parameter covariate relations that have been studied
also resulted in no improvement of the model fit. These were
serum bilirubin in non-renal elimination, body weight in every parameter (except those previously mentioned) or fu and plasma
protein levels in inter-compartment CL. Some authors have also
suggested that variations in the V1 of the drugs could be related
to the severity of the level of sickness.12 However, in our population
neither the APACHE II nor the SOFA scores influenced V1 in a relevant way.
Consequently, the initial high IIV estimated by the base models
was partially explained by the inclusion of covariates. However,
high variability was still estimated with the selected model,
mainly for piperacillin. The heterogeneity of ICU patients is well
known and it could lead to high IIV. Furthermore, a big limitation
in the present population study was the small cohort size, which
may prevent other covariates from being shown to be significant.
Regarding model evaluation, the relative standard errors were
,40% for CLNR and lower for the rest of the parameters. The robustness of estimation was demonstrated by the bootstrap method. In
this study, the pharmacokinetic parameters were consistent with
the values estimated by pharmacokinetic approaches previously
published,1 3,11,14,15 depending on the residual renal function
and the status of the patient. The suitability of the final models
was demonstrated by the performance of pvcVPCs, where simulations conducted with the final models supported the observed
data. The pvcVPC data values have been normalized by the prediction, which allows us to deal with subpopulations with very different individuals.10
Once a population pharmacokinetic model had been developed
for piperacillin and tazobactam, the pharmacokinetic/pharmacodynamic analysis was performed. Pharmacokinetic/pharmacodynamic analysis is useful to assess the probable output of a
certain dosing regimen and allow recommendations on prescription to maximize the probability of success of the therapy. For
this purpose, some simulations were run with different values for
CLCR (10, 50 and 100 mL/min), the inter-dose interval (4, 6 and
8 h) and the duration of perfusion. The pharmacokinetic/pharmacodynamic target related to the success of therapy with b-lactams
is to maintain concentrations of unbound antibiotic above the MIC
for the microorganism responsible for the infection, as they are
antibiotics with a time-dependent activity pattern.21 Since the

185

Downloaded from http://jac.oxfordjournals.org/ by guest on December 6, 2015

JAC

Asn-Prieto et al.

(a)

(d)
100%

PTA (%)

80%
60%

MIC (mg/L)
1

16

4 g q4 h

100

100

99

91

60

4 g q6 h

99

96

82

49

16

4 g q8 h

92

76

46

17

40%
20%
64
12
8

32

8
16

0.
5

0.
12

5
0.
25

0%

(e)
PTA (%)

100%

MIC (mg/L)

80%

16

32

60%

4 g q4 h

100

100

100

93

63

4 g q6 h

100

99

91

62

22

4 g q8 h

98

91

66

29

40%
20%
8

12

64

16
32

0.

12
5
0.
25
0.
5

0%

(c)

(f)
PTA (%)

100%

60%

MIC (mg/L)
4

16

32

64

4 g q6 h

100

100

100

92

55

4 g q8 h

100

100

96

72

27

80%

40%
20%
64
12
8

32

8
16

5
0.

25

0.

0.

12

0%

MIC (mg/L)
Figure 3. PTA for piperacillin administered every 4 h (open squares), 6 h (filled circles) or 8 h (open circles) to patients with CLCR values of 100 mL/min (a),
50 mL/min (b) and 10 mL/min (c). The dashed line corresponds to a PTA value of 90%, which is related to high success probabilities. PTA values for
virtual patients with CLCR of 100 mL/min (d), 50 mL/min (e) and 10 mL/min (f) are detailed in the adjacent tables. q4 h, administered every 4 h; q6 h,
administered every 6 h; q8 h, administered every 8 h.

subjects were critically ill patients, unbound drug levels should be

above the MIC for the microorganism responsible for the infection
during the entire dosing interval (fT.MIC 100%).12,13 When simulating the pharmacokinetic profiles in virtual patients with a high
rate of renal excretion (CLCR 100 mL/min) receiving 4/0.5 g of
piperacillin/tazobactam in a short infusion, noticeable differences
in the PTA were found among the dosing intervals depending on
the MIC (Figure 3). High probabilities could be expected at
1 mg/L giving the drug every 8 h. When shortening the dosing
interval, high probabilities of success are achieved at larger MICs.
In fact, if a patient with high CLCR is infected by a microorganism
for which the MIC is 2 mg/L, administration of piperacillin/tazobactam every 6 h is needed, whereas if the MIC for the pathogen is
8 mg/L a 4 h dosing interval must be used to guarantee the
outcome of the therapy. Regarding virtual patients undergoing
moderate renal failure, infections by microorganisms for which
the MICs are up to 16 mg/L could be treated, but just giving 4 g
of piperacillin every 4 h (pathogens for which MICs are 4 and

186

8 mg/L could also be treated with 4 g of piperacillin every 8 and

6 h, respectively). Finally, in the case of virtual patients with low
CLCR values, 4 g of piperacillin every 6 h could be used for infections
caused by microorganisms for which the MICs are up to 32 mg/L
(4 g every 8 h gave high probabilities of success with MICs
16 mg/L).
Some authors have commented that the piperacillin/tazobactam
dosage must be lowered to 2/0.25 g every 6 h in patients receiving
CRRT.22 Based on our findings, following this recommendation
patients with moderate or good renal function receiving CRRT
would be underdosed. In contrast, Gilbert et al.16 recommend increasing the dosing interval to 8 h in anuric patients instead of lowering the dosage, which is in line with our results (high PTAs up to an
MIC of 16 mg/L with 4 g of piperacillin every 8 h).
Our analyses confirm that a 20 min infusion of 4 g of piperacillin
is able to cover infections caused by pathogens for which the
MICs are low (,8 mg/L). Pathogens for which MICs are higher
could be treated with short-infusion piperacillin/tazobactam, but

Downloaded from http://jac.oxfordjournals.org/ by guest on December 6, 2015

PTA (fT>MIC=100%)

(b)

JAC

Piperacillin/tazobactam population PK/PD during CRRT

(a)

(c)

100%

PTA (%)

MIC (mg/L)
2

16

32

CI 24 g q24 h

100

100

100

99

80

CI 16 g q24 h

100

100

100

92

51

CI 12 g q24 h

100

100

99

80

29

EI 4 g q6 h

100

99

89

52

13

EI 4 g q8 h

97

82

47

14

80%
60%
40%
20%
64
12
8

32

16

0.
12
5
0.
25
0.
5

PTA (fT>MIC=100%)

0%

(b)

(d)
PTA (%)

100%

16

32

CI 24 g q24 h

100

100

100

100

94

CI 16 g q24 h

100

100

100

98

78

CI 12 g q24 h

100

100

100

94

59

EI 4 g q6 h

100

100

99

89

49

EI 4 g q8 h

100

99

89

55

16

40%
20%
64
12
8

32

8
16

0.

12
5
0.
25
0.
5

0%

MIC (mg/L)

Figure 4. PTA for piperacillin administered in a continuous infusion of 24 g every 24 h (open triangles), a continuous infusion of 16 g every 24 h (open
circles), a continuous infusion of 12 g every 24 h (open squares), an extended infusion of 4 g every 6 h (filled circles) and an extended infusion of 4 g
every 8 h (filled squares) to patients with CLCR values of 100 mL/min (a) and 50 mL/min (b). The dashed line corresponds to a PTA value of 90%, which
is related to high success probabilities. PTA values for virtual patients with CLCR of 100 mL/min (c) and 50 mL/min (d) are detailed in the adjacent
tables. CI, continuous infusion; EI, extended infusion (4 h); q24 h, administered every 24 h; q6 h, administered every 6 h; q8 h, administered every 8 h.

prescribing the appropriate dosing regimen according to the simulations. Actually, for many of the clinical strains isolated from the
patients the MICs values were 16 mg/L. According to the results
obtained in our simulations, the administration of 4 g piperacillin
in a short infusion provides a low probability of success against
MICs of 16 mg/L if the patient exhibits a high CLCR (100 mL/min)
and if the patient has a CLCR ,50 mL/min only guarantees
success when the antimicrobial is given every 4 h. In order to
assess infections with higher MICs and provide a better approach,
simulations of extended infusion (4 h infusion time) of piperacillin/tazobactam 4/0.5 g every 6 or 8 h and continuous infusion of
12, 16 or 24 g every 24 h were performed in virtual patients
with CLCR values .50 mL/min (Figure 4). In line with other
studies where intermittent infusion, extended infusion and continuous infusion were compared,12,23,24 we observed better probabilities when prolonging the infusion time. For instance, the
administration of 12 g/day as a continuous infusion in patients
with a CLCR of 50 mL/min provides high probabilities of success
against bacteria for which the MIC values are up to 16 mg/L,
whereas 24 g/day is needed if drugs are administered as a short
infusion. Moreover, the administration of 24 g/day by continuous
infusion achieves PTAs .90% against MICs up to 32 mg/L.
However, extended perfusion does not achieve better results
than short infusion in this group of patients. When simulations

were carried out for patients with CLCR of 100 mL/min we

observed that the administration of 12 and 16 g/day by continuous infusion also provided high PTAs against microorganisms for
which the MICs were 8 and 16 mg/L, respectively. Meanwhile,
the administration of piperacillin by intermittent infusion seems
to be unable to guarantee adequate concentrations against
pathogens for which the MIC is 16 mg/L, and at least 24 g/day
is needed when the MIC is 8 mg/L.
Specifically, the use of piperacillin/tazobactam for the treatment of infections caused by P. aeruginosa is controversial.
Four of the patients included in this study presented infection
by P. aeruginosa, with MICs of 16 mg/L in three patients and
64 mg/L in one subject. Recently, the CLSI lowered the piperacillin/tazobactam breakpoint for P. aeruginosa from an MIC of 64
to 16 mg/L.25 Consequently, researchers have recommended
extended and continuous infusion.26,27 Nevertheless, the prolongation of infusion time must be revised when renal function
is affected, as the probabilities of success at high MICs are high
enough with 20 min of infusion (PTAs .90% at an MIC of
16 mg/L were obtained for virtual patients with acute renal
failure, and also for those with moderate renal failure administered piperacillin every 4 h), although a reduction in the total
daily dose must also be considered when extended or continuous infusions are used.

187

Downloaded from http://jac.oxfordjournals.org/ by guest on December 6, 2015

80%
60%

MIC (mg/L)

Asn-Prieto et al.

10a

50a

100a

(b) 18

16

16

14

14

12

12
81 80

10
8
68 67

65

Clearance (L/h)

Clearance (L/h)

(a) 18

1.5 2

4 6 8

3 4

1 1.5 2
Quf (L/h)

1 1.5

2
0

36

72

Conclusions

35
33
15

14

14
8

26 32

14
7 11

4 6

1.5 2

1 1.5 2
Quf (L/h)

1 1.5

Figure 5. Influence of the different components of clearance and the

impact of effluent rate on total body clearance of piperacillin (a) and
tazobactam (b) depending on CLCR (10, 50 and 100 mL/min). Numbers
inside bars indicate clearance as a percentage of respective of total body
clearance. Black bars, CLR; grey bars, CLNR (it was varied based on the
distribution of the post hoc values for the respective subpopulation);
white bars, CLEC. aCLCR (mL/min).

Tazobactam concentration (mg/L)

these subpopulations. For the CLEC estimation, both Quf and fu

were varied based on the distribution shown in patients with moderate renal failure. As observed in other studies,28 a relative accumulation is also shown in this research (Figure 6), but it does not
lead to toxic concentrations. On the other hand, Mueller et al.28
also suggested that the concentration of tazobactam should be
.4 mg/L to ensure the effectiveness of the piperacillin/tazobactam combination. As Figure 6 shows, simulated levels of tazobactam are greater than this critical concentration for the entire dosing
interval and the alternation of tazobactam administration suggested by van der Werf et al.14 could result in a decrease in its
activity.

30
25
20
15
10

Population pharmacokinetic models have been developed for

piperacillin and tazobactam when administered to critically ill
patients with different degrees of renal function who are receiving
CRRT. The pharmacokinetics of both drugs was best described by
two-compartment models. Total body CL was the sum of CLNR, CLR
(affected by CLCR) and CLEC. Body weight was a significant covariate
for piperacillin V2 and the type of patient for tazobactam V1. This
model is useful in establishing dosing recommendations considering the residual CLCR of the patient. Taking into account that free
drug trough concentrations above the MIC (fT.MIC 100%) have
to be guaranteed in the treatment of critically ill patients with
b-lactam antimicrobials and considering the MIC values for the
organisms potentially responsible for infection, our population
model allowed us to predict PTA depending on the dose regimen
and the CLCR of the patient. The pharmacokinetic/pharmacodynamic analysis confirmed that piperacillin/tazobactam is able to
cover infections by pathogens for which the MIC values are up to
16 mg/L if the CLCR of the patient is 100 mL/min when continuous
perfusion is used. When the CLCR is 50 mL/min, piperacillin/tazobactam provides high probabilities of success against microorganisms for which the MICs are up to 32 mg/L, although 24 g/day in
continuous perfusion may be needed. For bacterial isolates for
which MICs are lower, administration by continuous perfusion
allows the total daily dose to be reduced in comparison with short
intermittent perfusion.

5
0
0

12 16 20 24 72 76 80 84 88 92 96
Time (h)

Figure 6. Results from the simulation of virtual patients with moderate

renal failure (CLCR 50 mL/min) that were administered 0.5 g of
tazobactam every 4 h. Median tazobactam plasma levels are shown for
the simulated septic population (dashed line) and the polytraumatized
population (continuous line). The dotted line corresponds to a
concentration of 4 mg/L.

Funding
This work was supported by the Department of Education, Universities and
Research (IT341-10), Basque Government, Spain. E. A.-P. has been awarded
a research grant (grant number AE BFI-2010-116) from the Basque Government.

Transparency declarations
None to declare.

Finally, some simulations were carried out for tazobactam in

order to determine whether this b-lactamase inhibitor is accumulated in patients undergoing CRRT, as has been previously suggested.14,22 A thousand virtual patients with moderate renal
failure (CLCR 50 mL/min) that were given 0.5 g of tazobactam
every 4 h were simulated. Both septic and polytraumatized
patients were simulated in order to detect differences between

188

References
1 Landersdorfer CB, Bulitta JB, Kirkpatrick CM et al. Population
pharmacokinetics of piperacillin at two dose levels: influence of nonlinear
pharmacokinetics on the pharmacodynamic profile. Antimicrob Agents
Chemother 2012; 56: 571523.

Downloaded from http://jac.oxfordjournals.org/ by guest on December 6, 2015

36 34
0 20 26

78 75

38
16 16

86

4
27
26 27

100a

88

44 40

50a

10

10a

Piperacillin/tazobactam population PK/PD during CRRT

JAC

2 Arzuaga A, Maynar J, Gascon AR et al. Influence of renal function on the

pharmacokinetics of piperacillin/tazobactam in intensive care unit patients
during continuous venovenous hemofiltration. J Clin Pharmacol 2005; 45:
16876.

17 Vinks AA, Den Hollander JG, Overbeek SE et al. Population

pharmacokinetic analysis of nonlinear behavior of piperacillin during
intermittent or continuous infusion in patients with cystic fibrosis.
Antimicrob Agents Chemother 2003; 47: 5417.

3 Li C, Kuti JL, Nightingale CH et al. Population pharmacokinetics and

pharmacodynamics of piperacillin/tazobactam in patients with complicated
intra-abdominal infection. J Antimicrob Chemother 2005; 56: 38895.

18 Lodise TP Jr, Lomaestro B, Rodvold KA et al. Pharmacodynamic profiling

of piperacillin in the presence of tazobactam in patients through the use of
population pharmacokinetic models and Monte Carlo simulation.
Antimicrob Agents Chemother 2004; 48: 4718 24.

4 Arzuaga A, Isla A, Gascon AR et al. Elimination of piperacillin and

tazobactam by renal replacement therapies with AN69 and polysulfone
hemofilters: evaluation of the sieving coefficient. Blood Purif 2006; 24:
34754.
5 Cockcroft DW, Gault MH. Prediction of creatinine clearance from serum
creatinine. Nephron 1976; 16: 31 41.
6 Schetz M. Drug dosing in continuous renal replacement therapy: general
rules. Curr Opin Crit Care 2007; 13: 64551.

8 Wahlby U, Thomson AH, Milligan PA et al. Models for time-varying

covariates in population pharmacokinetic-pharmacodynamic analysis. Br
J Clin Pharmacol 2004; 58: 36777.
9 Lindbom L, Pihlgren P, Jonsson N. PsN-Toolkita collection of computer
intensive statistical methods for non-linear mixed effect modeling using
NONMEM. Comput Methods Programs Biomed 2005; 79: 24157.
10 Bergstrand M, Hooker AC, Wallin JE et al. Prediction-corrected visual
predictive checks for diagnosing nonlinear mixed-effects models. AAPS J
2011; 13: 14351.
11 Bauer SR, Salem C, Connor MJ Jr et al. Pharmacokinetics and
pharmacodynamics of piperacillin-tazobactam in 42 patients treated
with concomitant CRRT. Clin J Am Soc Nephrol 2012; 7: 4527.
12 Roberts JA, Kirkpatrick CM, Roberts MS et al. First-dose and steady-state
population pharmacokinetics and pharmacodynamics of piperacillin by
continuous or intermittent dosing in critically ill patients with sepsis. Int J
Antimicrob Agents 2010; 35: 15663.
13 Valtonen M, Tiula E, Takkunen O et al. Elimination of the piperacillin/
tazobactam combination during continuous venovenous haemofiltration
and haemodiafiltration in patients with acute renal failure. J Antimicrob
Chemother 2001; 48: 881 5.
14 van der Werf TS, Mulder PO, Zijlstra JG et al. Pharmacokinetics of
piperacillin and tazobactam in critically ill patients with renal failure,
treated with continuous veno-venous hemofiltration (CVVH). Intensive
Care Med 1997; 23: 8737.

20 Isla A, Rodriguez-Gascon A, Troconiz IFet al. Population pharmacokinetics

of meropenem in critically ill patients undergoing continuous renal
replacement therapy. Clin Pharmacokinet 2008; 47: 17380.
21 Hyatt JM, McKinnon PS, Zimmer GS et al. The importance of pharmacokinetic/pharmacodynamic surrogate markers to outcome. Focus on
antibacterial agents. Clin Pharmacokinet 1995; 28: 14360.
22 Trotman RL, Williamson JC, Shoemaker DM et al. Antibiotic dosing in
critically ill adult patients receiving continuous renal replacement
therapy. Clin Infect Dis 2005; 41: 1159 66.
23 Felton TW, Hope WW, Lomaestro BM et al. Population pharmacokinetics
of extended-infusion piperacillin-tazobactam in hospitalized patients
with nosocomial infections. Antimicrob Agents Chemother 2012; 56:
4087 94.
24 Roberts JA, Paratz J, Paratz E et al. Continuous infusion of b-lactam
antibiotics in severe infections: a review of its role. Int J Antimicrob Agents
2007; 30: 118.
25 Clinical and Laboratory Standards Institute. Performance Standards
for Antimicrobial Susceptibility Testing: Twenty-second Informational
Supplement M100-S22. CLSI, Wayne, PA, USA, 2012.
26 Kuti JL, Nightingale CH, Quintiliani R et al. Pharmacodynamic profiling of
continuously infused piperacillin/tazobactam against Pseudomonas
aeruginosa using Monte Carlo analysis. Diagn Microbiol Infect Dis 2002; 44:
517.
27 Lee GC, Liou H, Yee R et al. Outcomes of extended-infusion
piperacillin-tazobactam: a retrospective analysis of critically ill patients.
Clin Ther 2012; 34: 2297 300.
28 Mueller SC, Majcher-Peszynska J, Hickstein H et al. Pharmacokinetics of
piperacillin-tazobactam in anuric intensive care patients during continuous venovenous hemodialysis. Antimicrob Agents Chemother 2002; 46:
1557 60.

15 Bulitta JB, Duffull SB, Kinzig-Schippers M et al. Systematic comparison of

the population pharmacokinetics and pharmacodynamics of piperacillin
in cystic fibrosis patients and healthy volunteers. Antimicrob Agents
Chemother 2007; 51: 2497 507.

29 Vincent JL, de Mendonca A, Cantraine F et al. Use of the SOFA score

to assess the incidence of organ dysfunction/failure in intensive care
units: results of a multicenter, prospective study. Working group on
sepsis-related problems of the European Society of Intensive Care
Medicine. Crit Care Med 1998; 26: 1793 800.

16 Gilbert B, Robbins P, Livornese LL Jr. Use of antibacterial agents in renal

failure. Infect Dis Clin North Am 2009; 23: 899924.

30 Knaus WA, Draper EA, Wagner DP et al. APACHE II: a severity of disease
classification system. Crit Care Med 1985; 13: 81829.

189

Downloaded from http://jac.oxfordjournals.org/ by guest on December 6, 2015

7 Arzuaga A, Isla A, Gascon AR et al. Quantitation and stability of piperacillin

and tazobactam in plasma and ultrafiltrate from patients undergoing
continuous venovenous hemofiltration by HPLC. Biomed Chromatogr
2005; 19: 5708.

19 Bulitta JB, Kinzig M, Jakob V et al. Nonlinear pharmacokinetics of

piperacillin in healthy volunteers-implications for optimal dosage regimens.
Br J Clin Pharmacol 2010; 70: 68293.