Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Center for the Advancement of Research and Education in Infectious Diseases, The University of Texas at
Austin College of Pharmacy and Pharmacotherapy Research and Education Center, Austin, Texas; and the
University of Texas Health Science Center at San Antonio, San Antonio, Texas
ABSTRACT
Objective: This study compared the pharmacokineticl
pharmacodynamic (PK/PD) properties of piperacillinl
tazobactam (PTZ) combination treatment with those
of piperacillin (PIP) monotherapy against clinical gramnegative pulmonary isolates from US patients treated in
intensive care units.
Methods: Computer modeling was used to integrate
national in vitro microbiologic data from 2002 with
pharmacokinetic data from published studies in healthy
volunteers. PTZ (3.375 g q4h, 3.375 g q6h, 4.5 g q6h,
and 4.5 g q8h) and PIP (3 g q4h, 3 g q6h, 4 g q6h, and
4 g q8h) were modeled using Monte Carlo simulations.
The cumulative fraction of response (CFR) was determined for percentage of time that the free serum concentration remained above the MIC of ::::30% (bacteriostatic) and ::::50% (bactericidal). Because simulated
comparisons with an artificially derived sample size were
used, statistical methods were not applied.
Results: Overall, 2584 gram-negative pulmonary
isolates were evaluated, including Enterobacteriaceae
(n = 1430), Pseudomonas aeruginosa (n = 799),
Acinetobacter baumannii (n = 179), and "other" (n =
176). The percents susceptible with PTZ and PIP were
as follows: Enterobacteriaceae, 86% and 66%, respectively; P aeruginosa_. 89% and 84%; and A baumannii_.
47% and 34%. CFR rates with PTZ were numerically
higher than those with PIP against Enterobacteriaceae
(ranges, 86%-89% and 66%-73%, respectively) and
A bazt1nannii (47%-53% and 33%-42%), but not
against P aeruginosa (79%-84% and 75%-81 %).
Conclusion: Results from PKlPD models with Monte
Carlo simulation suggested that susceptibility differences
among these selected gram-negative isolates collected in
December 2008
2002 may be of sufficient magnitude to result in notable PK/PD differences between PTZ and PIP. (Clin The]:
2008;30:2335-2341) 2008 Excerpta Medica Inc.
Key words: Monte Carlo simulation, piperacillinl
tazobactam, aerobic gram-negative bacteria, pulmonary disease, infectious disease.
INTRODUCTION
The primary resistance mechanism of gram-negative
bacteria against ~-lactam antibiotics is the production
of ~-lactamase enzymes. 1 These enzymes cleave the
~-lactam ring, thereby rendering ~-lactam antibiotics
(eg, piperacillin [PIP]) ineffective. ~-Lactamase inhibitors (eg, tazobactam) counteract the ~-lactamase enzyme and protect the ~-lactam antibiotic from degradation. In studies, combinations of a ~-lactam plus a
~-lactamase inhibitor (eg, piperacillin/tazobactam [PTZ])
have been found to be more active compared with
monotherapy with ~-lactam antibiotics against bacteria that produce ~-lactamase.2
Over the past decade, the pharmaceutical industry
has begun to monitor antibiotic resistance, including
that which results from increased production of
~-lactamase. These programs provide valuable information regarding bacterial epidemiology and emerging resistance. One such surveillance program is the
Intensive Care Unit Surveillance System (ISS), developed by Merck & Co, Inc. (Rahway, New Jersey). 3,4
Accepted for publicatIOn October 31,2008
dOl:1 0_1 016!J-cIinthera_2008_12_009
0149-2918/$32_00
2335
Clinical Therapeutics
The study of pharmacokinetic and pharmacodynamic (PK/PD) properties involves the integration of
pharmacokinetic properties and microbiologic activity
and is an important consideration when evaluating
antimicrobial alternatives. 5-7 The PK/PD profile of each
antimicrobial can be classified as one of the following:
Cma)MIC, AUC/MIC, or the percentage of time that
the free serum concentration remains above the MIC
(f% T>Mld 8- 1O For aminoglycosides, the literature suggests that Cma)MIC is the best predictor of clinical
outcome, whereas for fluoroquinolones, the literature
supports free AUC/MIC as the best predictor of clinical outcome. 8- 1O For ~-lactams, the literature supports
T>MIC as the best predictor of clinical outcome. 8- 1O PK/
PD profiles can be compared based on the mean serum
concentration and a single MIC value, usually MIC go .
However, this approach does not take into account the
variability of the pharmacokinetics or the distribution
of MICs; hence, this single-point PK/PD analysis can
indicate what is possible but not what is probable. ll
Monte Carlo simulation is a method that integrates the
pharmacokinetic and microbiologic data to determine
the probability of achieving a specific parameter.
Given our understanding of PK/PD principles, it
follows that the enhanced in vitro activity of PTZ is
relevant only if the following are true: (1) ~-lactamase
producing bacteria are prevalent, and (2) the presence
of ~-lactamase results in susceptibility changes of sufficient magnitude to compromise the PK/PD properties of PIP. The present study used PK/PD modeling to
predict the potential PK/PD benefits of PTZ compared
with PIP for the treatment gram-negative pulmonary
infections in the intensive care unit (ICU).
Table I. Published pharmacokinetic (PK) properties of piperacillin/tazobactam (PTZ) and piperacillin (PIP)
monotherapy in healthy volunteers. 13 * Data are mean (sD).
PTZ
PK Parameter
PIP
3 g q4h, q6h t
4 g q6h, q8h t
11.2(2.1)
0.76 (0.11)
10.5 (1.4)
0.75 (0.10)
11.2(2.1)
0.76 (0.11)
10.5 (1.4)
0.75 (0.10)
2336
Volume 30 Number 12
ducted for each regimen-bacteria combination. Pharmacokinetic and microbiologic data were integrated
according to a I-compartment IV bolus model. IS Vss
and tl/2~ were varied according to logarithm-normal
distributions, which were constructed using the published mean (SD) values. Protein binding was varied
according to a uniform distribution (30% [SD 10%]).
In contrast, the actual MIC values from the microbiologic data were used to construct discrete distributions. The cumulative fraction of response (CFR) was
determined for f% T >MIC values of ::::30% (bacteriostatic) and ::::50% (bactericidal).l0
Because this study involved simulated patients, the
relationships described herein should be viewed as
qualitative as opposed to quantitative. With simulations such as these, the sample size is artificially chosen; in this case, we chose 10,000 simulations by
convention. Because the size of the P value is largely
dependent on the chosen sample size in simulated
comparisons, we believed it would not have been valid
to apply statistical methods to these simulated comparisons. We also point to similar studies that have
not applied statistics to these simulations. 16,17
RESULTS
Overall, the 2002 ISS database included susceptibility
data from 2584 gram-negative pulmonary isolates
(Enterobacteriaceae, 1430; P aeruginosa_. 799; A balt11lannii. 179; and "other," 176). Table II depicts the in
vitro activity of PTZ and PIP against these clinical
isolates. Against Enterobacteriaceae, the MIC so and
MIC gO were similar with PTZ (4/4 and 64/4 rg/mL,
respectively) and PIP (4 and 128 rg/mL); however,
the percent susceptible was 20% higher with PTZ
than PIP (86% vs 66%). For P aeruginosa, values
were similar between PTZ and PIP in terms of MIC so
(4/4 vs 4 rg/mL), MIC gO (128/4 vs 128 rg/mL), and
percent susceptible (89% vs 84%). In contrast, for
A bazt1nannii, PTZ had a lower MIC so than PIP (32/4 vs
128 flg/mL) and a higher percent susceptible than PIP
(47% vs 34%), whereas the MIC gO was similar between the 2 antibiotics (128/4 vs 128 rg/mL).
Table III and the figure show the PK/PD results for
both the bacteriostatic (f% T >MIC ::::30%) and bactericidal (f% T >MIC ::::50%) targets. For the bacteriostatic
target, CFR rates were higher with PTZ than PIP
against Enterobacteriaceae (ranges, 86%-89% and
66%-73 %, respectively) and A balt11lannii (47%-53 %
and 33 %-42 %) but similar against P aeruginosa
December 2008
2337
Clinical Therapeutics
Table II. Frequency distribution of MICs of piperacillin/tazobactam (PTZ) and piperacillin (PIP) monotherapy
for gram-negative pulmonary isolates from the 2002 Intensive Care Unit Surveillance System database.
Enterobacteriaceae
(n = 1430)
Pseudomonas aeruginosa
(n = 799)
Acinetobacter baumannii
(n = 179)
MIC, I-lg/mL
PTZ
PIP
PTZ
PIP
PTZ
PIP
:C:;4
0.77*
0.50*
0.57*
0.51 *
0.32
0.07
0.84
0.60
0.72
0.67
0.42
0.21
16
0.86
0.66
0.80
0.77
0.47
0.34
32
0.89
0.73
0.84
0.81
0.50*
0.42
64
0.93t
0.78
0.89
0.84
0.69
0.45
:::128
1.00
1.00t
1.00t
1.00t
1.00t
1.00*t
%5*
86
66
89
84
47
34
%S ~ percent susceptible.
*MIC so '
t MIC 9o '
f Percent susceptible at the current susceptibility breakpoints of the Clinical and Laboratory Standards Institute for PTZ
and PIP: Enterobacteriaceae (~16/4 and ~16 IJg/mL), P aerugtnosa (~64/4 and ~64 IJg/mL), and A baumanntt (~16/4 and
~16IJg/mL).12
Table III. Pulmonary pharmacokinetic/pharmacodynamic data of piperacillin/tazobactam (PTZ) and piperacillin (PIP) monotherapy against gram-negative pulmonary isolates from patients hospitalized in the intensive care unit.* Values are %.
PTZ
PIP
4.5 g
3.375 g
Bacteria/ Rate
3g
4g
q4h
q6h
q6h
q8h
q4h
q6h
q6h
q8h
Enterobacteriaceae
Bacteriostatic
Bactericidal
89
85
87
68
88
83
86
54
73
65
67
47
71
60
66
37
Pseudomonas aeruginosa
Bacteriostatic
Bactericidal
84
78
80
56
84
72
79
42
81
74
76
51
80
68
75
38
Acinetobacter baumannii
Bacteriostatic
Bactericidal
53
46
48
33
50
42
47
24
42
32
35
13
40
24
33
8
*The cumulative fraction of response was determined for percentage of time that the free serum concentration remained
above the MIC of:::30% (bacteriostatic) and :::50% (bacteriCidal).
2338
Volume 30 Number 12
100
90
80
g
ci
LL
70
60
50
40
30
20
10
a~----,--------,-----,-----,----,---~,------~=::;;~~~~~
10
20
30
40
50
60
70
80
90
100
a
B
100
90
80
70
g
ci
LL
60
50
40
30
20
10
o--.L-,---------r------r---r--------r----;-------,r---~==~""""~~"""""=Q
10
20
30
40
50
60
70
80
90
100
100
90
80
70
60
ci
50
LL
40
30
20
10
a~-_____,_------r--___,_-_,__-~~~~~~~~::::::::s
10
20
30
40
50
60
70
80
90
100
a
December 2008
2339
Clinical Therapeutics
sired, health care providers should select the more intensive PTZ and PIP regimens for ICU patients with
gram-negative bacterial infections.
This study had some potential limitations regarding the microbiological data, pharmacokinetic data,
and the design of the PK/PD models. The 2002 ISS
database is a rich source of microbiological data that
focuses on ICU patients. It is unique because it allows
subgroup analyses by culture site 18 ; however, the database was not developed for PK/PD modeling, so in
vitro activity was evaluated only over a narrow range
of MICs from 4 to 128 rg/mL. As a result, more than
half of the MICs of PTZ and PIP against Enterobacteriaceae and P aeruginosa were assigned a MIC of
4 rg/mL when the actual MIC was at or below the
limit of detection (ie, 4 rg/mL). Lower MICs would
have caused the CFR rates to be higher. In contrast,
for A balt11lannii, 31 % of isolates for PTZ and 55%
of isolates for PIP were at or above the highest limit
of detection (ie, 128 rg/mL). Higher MICs would
have resulted in lower CFRs for each of these antibiotics against A balt11lannii. Another limitation about the
microbiological data is that they were collected in
2002, and their clinical relevance to the current year
is unknown.
Pharmacokinetic data were derived from healthy
volunteers and may not accurately represent the values observed among critically ill patients. Because
most ~-lactams are renally eliminated, compromised
renal function may result in elevated drug concentrations and improved CFR rates. In addition, the pharmacokinetic parameters were measured in serum, but
the results were extrapolated to patients with pulmonary infections. Pulmonary pharmacokinetics would
be preferable, but to our knowledge, they are not
available in a format that can be used for PK/PD modeling. Finally, pharmacokinetic data studies are typically quite small (<10 patients) and the one that was
used for the pharmacokinetics in this study was no
exception; therefore, it is important to realize that although the microbiological data in this study were
derived from >2500 MICs, the pharmacokinetic data
were derived from only a handful of patients, as is the
case for most PK/PD simulation studies. 19-21
The PK/PD relationships were based on animal
studies, and most have not been confirmed in human
studies. 8- 1O This study used a I-compartment PK/PD
model that did not account for the % T >MIC contributed by the time of infusion. Recent investigations
2340
have suggested that the infusion time can be prolonged to enhance the % T >MIC' 22-24 The PK/PD impact
would be minimal for short infusions (:c:;30 minutes),
but prolonged infusions would enhance the ability of a
~-lactam to achieve its PK/PD targets. In defense of the
methods used in this study, any PK/PD benefit achieved
by prolonged infusions should be similar for PIP and
PTZ. Finally, the focus of this PK/PD model was efficacy, and simulations regarding tolerability were not
attempted. Therefore, although more intensive dosing
regimens might improve PK/PD, it is uncertain whether
this improvement would be outweighed by an increased risk for adverse events.
CONCLUSIONS
Based on the results from the present comparisons
of the in vitro activities of PTZ and PIP in the present
study, it can be deduced that many gram-negative
pulmonary isolates collected from US ICUs in 2002
produced ~-lactamase. ~-Lactamase production appeared to be more common among Enterobacteriaceae
and A balt11lannii than among P aeruginosa_. thereby
resulting in improved in vitro susceptibilities for PTZ
than for PIP against Enterobacteriaceae and A balt11lannii. Furthermore, PK/PD models with Monte
Carlo simulation suggested that the reductions in susceptibility for PIP as compared with PTZ may be of
sufficient magnitude to result in notable differences in
the PK/PD profiles of these antibiotics.
ACKNOWLEDGM ENTS
This research was sponsored by Wyeth Pharmaceuticals.
Dr. Burgess has received educational grants, honoraria,
and research grants from Wyeth and serves as a consultant for Wyeth. The sponsor did not have any role in the
design, conduct, analysis, or publication of this study.
The authors thank Brittany Makos and Christine
Oramasionwu, PharmD, for their technical assistance
with this manuscript.
REFERENCES
1. Tenover Fe. Mechanisms of antimicrobial resistance m
bactena. Am J Med. 2006;119(5uppl 1):53-51 0; discussion 562-570.
2. Fass RJ, Pnor RB. Comparative m Vitro activities of pi peracil Ii ntazobactam and tlcarcdlm-c1avulanate. Antlmlcrob Agents
Chemother. 1989;33: 1268-1274.
3. Itokazu G5, QumnJP, Bell-Dixon C, et al. Antimicrobial resIstance rates among aerobic gram-negative bacilli recovered
Volume 30 Number 12
4.
5.
6.
7.
77.
8. Craig WA. Interrelationship between pharmacokinetics and pharmacodynamics In determining dosage regimens for broad-spectrum
cephalosporlns. Dlagn Mlcroblol Infect DIs. 1995;22:89-96.
9. Craig WA. Pharmacoklnetlc/pharmacodynamlc parameters: Rationale for
antibacterial dosing of mice and men.
Cltn Infect DIs. 1998;26:1-10.
10. Orusano GL. Antimicrobial pharmacodynamics: Critical interactions
of ,bug and drug'. Nat Rev Mlcroblol.
October 7, 2008.
15. Turnldge jO. The pharmacodynamICS of beta-Iactams. Clm Infect DIs.
1998;27:10-22.
16. OeRyke CA, Kutl j L, Nlcolau OP.
Reevaluation of current susceptibility breakpoints for Gram-negative
rods based on pharmacodynamic
assessment. Dlagn Mlcroblol Infect
DIs. 2007;58:337-344.
17. Burgess OS, Frel CR, LewIs II js, et
al. The contribution of pharmacokinetic-pharmacodynamic modelling with Monte Carlo simulation to
the development of susceptibility
breakpoints for Neisseria memngltldls.
Clm Mlcroblol Infect. 2007;13:3339.
18. Frel CR, Hampton sL, Burgess OS.
Influence of culture site-specific
MIC distributions on the pharmacoklnetlc and pharmacodynamic
properties of plperacdlin/tazobactam
2004;2:289-300.
11. White RL. What In vitro models of
infection can and cannot do. Pharmacotherapy. 2001 ;21 :2925-301 S.
December 2008