Clinical Therapeutics/Volume 29, Number 5, 2007
Empiric Therapy for Secondary Peritonitis: A
Pharmacodynamic Analysis of Cefepime, Ceftazidime,
Ceftriaxone, Imipenem, Levofloxacin, Piperacillin/
Tazobactam, and Tigecycline Using Monte Carlo Simulation
KathrynJ. Eagye, MPH1;Joseph L. Kuti, PharmD1; Michael Dowzicky, MS2;and
David P. Nicolau, PharmD, FCCP1,3
1Center for Anti-Infective Research and Development, Hartford Hospital, Hartford, Connecticut; 2Wyeth
Research, Collegeville, Pennsylvania;and 3Division of Infectious Diseases, Hartford Hospital, Hartford,
Connecticut
ABSTRACT
Background: Inappropriate antibiotic therapy (ie,
the selection of an empiric agent without activity against
the responsible pathogen) of secondary peritonitis may
result in poor patient outcomes. The selection of an
appropriate agent can be challenging because of the
emerging resistance of target organisms to commonly
prescribed antibiotics.
Objective: The aim of this study was to perform a
pharmacodynamic analysis, using recent global sur-
veillance data, of commonly prescribed antibiotic agents
and a newer agent, tigecycline, indicated in 2005 for
the treatment of complicated intra-abdominal infec-
tions, to determine their probability for achieving mi-
crobiologic success against aerobic bacteria associated
with secondary peritonitis.
Methods: A 2-compartment model was construct-
ed using pharmacokinetic data from critically ill pa-
tients and global surveillance data on MIC distribu-
tions for microorganisms encountered in secondary
peritonitis. A Monte Carlo simulation of the modeled
data was performed to determine drug-appropriate
pharmacodynamic end points, including free-drug
time above the MIC, steady-state concentration above
the MIC, and AUC/MIC ratios. A cumulative fraction
of response (CFR) against aerobic bacteria involved
in secondary peritonitis was calculated for cefepime,
ceftazidime, ceftriaxone, imipenem, levofloxacin,
piperacillin/tazobactam, and tigecycline. A CFR
_>90% was considered microbiologic success. The fol-
lowing treatment regimens, administered as 30-minute
IV infusions, were examined: cefepime 1 and 2 g
q12h, ceftazidime 1 and 2 g qSh, ceftriaxone 1 and
2 g q24h, imipenem 500 mg q6h, levofloxacin 750 mg
May2007
q24h, piperacillin/tazobactam 3.375 g q6h, and tige-
cycline 50 mg q12h, after a loading dose of 100 rag.
Results: A CFR _>90% against nonenterococcal bac-
teria was predicted for imipenem 500 mg q6h (96.8 %),
cefepime 2 and 1 g q12h (95.3% and 92.4%, respec-
tively), ceftazidime 2 g qSh (94.2%), and piperacillin/
tazobactam 3.375 g q6h (91.2%). A CFR of 84.5%
was predicted for tigecycline 50 mg q12h. Ceftriaxone
and levofloxacin were predicted to have a CFR <80%.
When enterococci were included in the model, the pre-
dicted CFRs for imipenem, piperacillin/tazobactam,
and tigecycline were 93.4%, 88.4%, and 86.7%,
respectively.
Conclusions: MIC distribution and pathogen preva-
lence strongly influence the likelihood of microbiologi-
cal success in secondary peritonitis; therefore, deci-
sions regarding empiric therapy should consider local
epidemiology. Using current global data, the following
regimens are adequate choices if Enterococcus is not
targeted: Combination therapy (with metronidazole)
using cefepime 1 g or 2 g q12h, or ceftazidime 2 g
qSh; or monotherapy with imipenem 500 mg q6h or
piperacillin-tazobactam 3.375 g q6h. When Entero-
coccus is included in the epidemiologic mix, imipen-
era, piperacillin/tazobactam, and tigecycline all appear
to be viable monotherapeutic choices. (Clin Ther.
2007;29:889-899) Copyright © 2007 Excerpta
Medica, Inc.
Acceptedfor pubficationMarch2, 200Z
doi:l 0.1016/j.clinthera.2007.05.018
0149-2918/532.00
Printed in the USA. Reproduction in whole or part is not permitted.
Copyright © 2007 Excerpta Medica, Inc.
889
 Clinical Therapeutics
Key words: secondary peritonitis, pharmacody-
namic profiling, Monte Carlo simulation, tigecycline.
INTRODUCTION
Secondary peritonitis is a serious condition that may
be caused by spillage of gut flora into the peritoneum
via a breach in the gut wall, necrotic tissue, or diver-
ticular disease; the resulting inflammatory response
and abscesses can be difficult to treat and are associ-
ated with mortality rates of up to 30%. 1-3 Compli-
cated intra-abdominal infections (clAIs) require surgical
drainage and d4bridement to eliminate the source of
infection, as well as adjuvant antibiotic therapy.3 Micro-
organisms resistant to many of the currently prescribed
antibiotics are increasingly encountered in clAIs.4
Inappropriate empiric antibiotic therapy (ie, the se-
lection of an empiric agent without activity against the
responsible pathogen) has been found to negatively in-
fluence patient outcomes. 5,6 Several recent studies have
found a significant relationship between inappropri-
ate therapy and lower rates of clinical success (53.4%
[95% CI, 41.1%-69.3%] vs 78.6% [95% CI, 73.6%-
83.9%] for appropriate therapy), (OR of clinical fail-
ure, 14.9 [P = 0.001]); longer lengths of stay (increase) of
3.38 days [95% CI, 3.38-3.39] vs appropriate therapy)
(16.5 days for clinical successes vs 9.0 days for failures
[P < 0.001]); and hospitalization costs (increase, ~1359
[95% CI, 1355-1362]).6-9 Guidelines developed by the
Infectious Diseases Society of America (IDSA) recom-
mend monotherapy with a carbapenem or ]3-1actarrd
]3-1actamase inhibitor or combination therapy with
metronidazole and a cephalosporin or quinolone as op-
tions for empiric therapy. 1° However, a meta-analysis
of 40 clinical trials in 5094 patients with secondary
peritonitis was unable to identify a difference in effica-
cy or toxicity among agents. 11 Tigecycline, which was
approved by the US Food and Drug Administration in
2005 for the treatment of clAIs and is the first agent
from the glycylcyclineclass to be marketed for this pur-
pose, has been found to have activity against many
Enterobacteriaceae, against Enterococcus spp (includ-
ing vancomycin-resistant enterococci [VRE]) and against
methicillin-resistant Staphylococcus aureus. 12,13 The
recommended therapeutic dose for tigecycline is 50 mg
q12h, preceded by a loading dose of 100 mg.
Because inappropriate choice of antibiotic regimens
has been associated with a 3.4-fold increase in the rate
of clinical failure and a mortality rate of -30%, the
890
stakes of empiric therapy selection are high. 3,9 In the
absence of clinical data revealing clear distinctions be-
tween therapeutic agents, other criteria must be used in
making an informed selection. Microbial factors, such
as the focus of infection and the local resistance profile
of relevant pathogens are important considerations,
as are the characteristics of individual antimicrobial
agents. The performance of a given compound can
vary--even when used against the same organism--
according to the ability of the patient to absorb, dis-
tribute, and eliminate the drug. Furthermore, agents
from different classes of antibiotics assert their antimi-
crobial effect in different ways; some may be time de-
pendent and others, concentration dependent. 14 The
pharmacokinetic and pharmacodynamic properties of
antimicrobial agents can be used as a basis for making
predictions about their success against a specific popu-
lation of organisms. Although clinical success depends
on many factors (including adequate source control
and patient pathophysiology), effective antimicrobial
therapies are important in the treatment of secondary
peritonitis, and increasing antimicrobial resistance may
alter the expected performance of commonly pre-
scribed agents. When considered in conjunction with
the latest susceptibility profile of relevant organisms,
pharmacokinetic and pharmacodynamic data can be
helpful in differentiating among therapeutic choices.
The investigators have previously employed these
techniques to examine therapeutic options in clMs. 15,16
The present study is an examination of the issue of
empiric therapy for such infections using updated
modeling techniques--including a 2-compartment
pharmacokinetic model instead of 1-compartment,
and pharmacokinetic parameters derived from criti-
cally ill patients instead of healthy volunteers--and a
recent and more comprehensive microbiological data
set. A new monotherapeutic agent, tigecycline, is also
included, as is a parallel analysis to include therapy
targeted to enterococci; these organisms were not
modeled in the previous analyses. The purpose of the
present study was to evaluate options for the empiric
treatment of secondary peritonitis, including tigecy-
cline, an agent newly marketed for this purpose, using
improved analytic methods and recent global surveil-
lance data on relevant pathogens.
MATERIALS AND METHODS
Pharmacokinetic/pharmacodynamic modeling and
Monte Carlo simulation techniques were used to es-
Volume 29 Number 5

timate the probability of antimicrobial compounds
achieving the maximum effective exposure against
organisms associated with cIAIs. Pharmacokinetic
models were constructed for these agents, and then a
Monte Carlo simulation was performed, using data
from critically ill patients, to incorporate patient
variability. The results from this simulation were
used to assess the pharmacodynamic exposure of
each agent against a relevant population of organ-
isms derived from 2 global surveillance studies. 17a8
The following antimicrobial regimens, administered
as 30-minute IV infusions, were examined: cefepime
1 and 2 g q12h, ceftazidime 1 and 2 g qSh, ceftriax-
one I and 2 g q24h, imipenem 500 mg q6h, levofloxa-
cin 750 mg q24h, piperacillin/tazobactam 3.375 g q6h,
and tigecycline 50 mg q12h, after a loading dose of
100 mg.
Pharmacokinetic Model
A 2-compartment, multiple-dose model was con-
structed to determine the 24-hour steady-state
concentration-time profile for each drug regimen.
Pharmacokinetic parameters, including total body
clearance, volume of the central compartment, and
microtransfer rate constants between the central and
peripheral compartments, were the input variables.
Estimates of unbound fraction for each drug were also
input variables.
Pharmacokinetic parameters for each compound
under study were obtained from previous pharmaco-
kinetic studies of antibiotic agents in patients who were
critically ill, except for unbound fraction, which was
derived from the package insert for each drug. The
mean and standard deviation for each of these param-
eters were reported in the previous studies. 19,2°
Covariance matrices were either reported in or were
calculated from data reported in these studies and
were applied to the pharmacokinetic variables during
the simulations.
Microbiology
Data on microbiologic pathogens involved in cIAIs
for 2004 to 2005 were obtained from the Tigecycline
Evaluation and Surveillance Trial (TEST Program) 17
and the SENTRY Antimicrobial Surveillance Program. is
It was necessary to use both of these databases in
order to obtain current data on MIC distributions
against all pathogens for all antimicrobial agents of
interest in this study. Data were extracted for organ-
May2007
K.J. Eagye et al.
isms that would be targeted in an empiric regimen for
the treatment of clAIs. 21 Anaerobic organisms were
excluded from the analysis. Since IDSA guidelines rec-
ommend monotherapy with agents that cover anaero-
bic bacteria or concomitant administration of metro-
nidazole with agents lacking anaerobic coverage, a
fair comparison of all antibiotic compounds should
include a parallel analysis of anaerobic coverage or
assume that the coverage provided by concomitant
administration of metronidazole would not be clini-
cally different from that provided by monotherapeutic
agents. Because this study already involves a paral-
lel analysis for Enterococcus--whose coverage is
controversial--including an additional parallel analy-
sis for anaerobes was deemed too unwieldy relative to
the utility of the information that might be gained.
The organisms of primary interest in this study were
Escherichia coli, Streptococcus spp (Streptococcus
pneumoniae or Group A streptococci), Klebsiella spp,
Pseudomonas aeruginosa, S aureus, Enterobacter spp,
Proteus mirabilis, and Acinetobacter spp. Data on
Enterococcus spp also were obtained for use in a par-
allel analysis. Data on the following isolates were de-
rived from the TEST Program: E coli (n = 2833),
Enterococcus spp (n = 1736), Klebsiella spp (n =
2750), S aureus (n = 1579), Enterobacter spp (n =
2548), and Acinetobacter spp (n = 1491). Data were
obtained from the SENTRY program for the fol-
lowing isolates: Streptococcus viridans (n = 669),
J3-hemolytic streptococci (n = 1764), P aeruginosa
(n = 3136), and P mirabilis (n = 742). Because data for
MIC distributions for ceftazidime against S aureus
and Enterococcus spp could not be obtained from the
TEST Program database, data for ceftriaxone MIC
distributions were substituted.
The resistance profile of each species was incorpo-
rated into the analysis as described under the section
Cumulative Fraction of Response. As part of the
analysis, species were weighted by their prevalence in
intra-abdominal infections; the prevalence rankings
are displayed in Table I. The rankings were based on
data derived from a double-blind, randomized, com-
parative, Phase III clinical trial by Solomkin et a121
comparing ertapenem with piperacillin/tazobactam
for the treatment of clAIs. This study was selected for
the timeliness and comprehensiveness of the data it
provided, as well as its appropriateness for use in a
comparative analysis with data from earlier studies by
the present investigators. 15,16
891
 Clinical Therapeutics
Table I. Prevalence of pathogens encountered in complicated intra-abdominal infections.*
Primary Analysis
Organism (Excluding Enterococcus)
Esckerickia coil
Streptococcus spp
Enterococcus s p p
Klebsiella spp
Pseudomonas aeruginosa
Staphylococcus aureus*
Enterobacter sp p
Proteus spp
Acinetobacter s p p
*Prevalence data were derived from a comparative Phase III clinical trial. 21
fPercentages may not total 100% because ofrounding.
tMethicillin-susceptible S aureus only.
Monte Carlo Simulation
A 5000-trial Monte Carlo simulation (Crystal Ball
2000, Descioneering, Denver, Colorado) was per-
formed using the pharmacokinetic model of each
antimicrobial regimen to determine the probability
of target attainment (PTA) profile for each agent.
P T A is the likelihood of a regimen meeting or exceed-
ing a predefined pharmacodynamic target at a given
MIC dilution. 22 In this study, the PTA was the pro-
portion of 5000 trials in each simulation during
which the target level of a pharmacodynamic index
was achieved at each MIC in doubling dilutions from
0.008 to 128 pg/mL. Indices were selected based on
the pharmacodynamic properties of each compound
(ie, time-dependent or concentration-dependent
killing). The pharmacodynamic index for bactericidal
target attainment was the concentration of flee (un-
bound) drug remaining above the MIC (f T>MIC)
for 40% of the dosing interval (40% f T>MIC) for
carbapenems and 50% (50% f T>MIC) for other
[3-1actam inhibitors and cephalosporins. 23,24 A total
AUC/MIC ratio >125 against gram-negative organ-
isms and >30 against gram-positive organisms was
considered bactericidal for levofloxacin2S,26; these
indices were >6.96 and >12.5, respectively, for tigecy-
cline. 27 Monte Carlo simulation was then performed
to determine the PTA profile of each antibiotic
regimen.
892
Prevalence (%)f
Parallel Analysis
(Including Enterococcus spp)
47.9 40.8
20.5 17.4
0 14.9
8.9 7.6
8.8 7.5
5.1 4.4
3.4 2.9
2.9 2.5
2.4 2.1
Cumulative Fraction of Response
The simulation results for PTAs were used to calcu-
late the cumulative fraction of response (CFR) for
each antibiotic regimen against pathogens involved in
secondary peritonitis. For each regimen, the PTA at
each MIC dilution was multiplied by the percentage
of isolates of each of the modeled pathogens found at
that dilution. The products of these multiplications
were summed to obtain the CFR for the regimen/
pathogen combination. 22
The CFRs for each antibiotic regimen/pathogen
combination were then weighted according to the
prevalence of the species in dAis; the weighted values
were then added to obtain the combined CFR of a regi-
men against all pathogens. The combined CFR repre-
sents the probability of an antibiotic regimen for
meeting or exceeding the desired pharmacodynamic
end point when used to target aerobic bacteria in the
treatment of secondary peritonitis. A CFR >90% was
considered the threshold of achieving a level of micro-
biologic efficacy that could be confidently relied on in
empiric therapy. 2s As a measure of significance, 9 5 %
CIs were calculated for each CFR, at a level of cz =
0.05, using the Newcombe-Wilson method without
correction for continuity. 29 However, since the num-
ber of simulation trials can be increased until signifi-
cance is reached, clinical significance should also be
considered.
Volume 29 Number 5

RES U LTS
CFRs obtained for each of the regimens against indi-
vidual pathogens and against all pathogen popu-
lations combined are presented in Tables II and
III, respectively. Agents with a predicted CFR _>90%
against nonenterococcal aerobic bacteria associated
with secondary peritonitis were cefepime, ceftazidime
(only at a dose of 2 g qSh), imipenem, and piperacillin/
tazobactam. The performance of these regimens re-
flected their excellent CFRs against the bacteria with
the highest prevalence in intra-abdominal infections
(and therefore in the simulation): E coli, Strepto-
coccus, and Klebsiella. None of these regimens dis-
played a CFR of less than -50% against any individual
organism, and most revealed moderate weakness in
but one or two of the less prevalent species modeled.
Tigecycline achieved an overall predicted CFR of
84.5%. Against individual species, it provided excel-
lent coverage of E coli, Streptococcus, and Staphylo-
coccus, with CFRs >95%. Coverage of Klebsiella,
Enterobacter, and Acinetobacter spp appeared to be
more moderate, with predicted CFRs ranging from
77% to 78%. As expected, tigecycline was found to
have a poor likelihood for achieving its target thresh-
old against P aeruginosa, with a predicted CFR of
1.9%. Poor coverage of P aeruginosa, which account-
ed for 8.8% of the modeled pathogens--as well as
moderate coverage of Klebsiella (at 8.9% of the mod-
eled pathogens)--strongly influenced the overall per-
formance of this drug.
Ceftriaxone and levofloxacin achieved predicted
CFRs <80%. This was expected in the case of ceftri-
axone, since it has been associated with low coverage
of P aeruginosa and S aureus. A predicted CFR of
74.7% against E coli was the strongest factor in low-
ering the overall results for levofloxacin.
When Enterococcus spp were included in the model,
the predicted CFRs of all agents were lower, except
for that of tigecycline, which increased from 84.5% to
86.7%. Predicted CFRs for imipenem were slightly
lower; however, the overall CFR remained >90% de-
spite a CFR of 74.2% against Enterococcus spp. The
increase in the overall predicted CFR for tigecycline
reflects its potent activity against this species: it
achieved a predicted CFR of 99.6% against Entero-
coccus spp, providing much greater coverage than the
next-best performing agent (imipenem). Predicted ex-
posures for cefepime and ceftazidime were substan-
tially lower, with values for predicted CFRs decreasing
May2007
K.J. Eagye et al.
by _>10%, principally reflecting the lack of activity
against enterococci. A similar decrease in predicted
CFRs was observed with ceftriaxone and levofloxa-
cin when enterococcal coverage was included in the
simulation.
DISCUSSION
The organisms typically present in secondary peritoni-
tis are proving to be increasingly resistant to our cur-
rent stock of antibiotics, which renders selection of an
appropriate empiric agent a challenge. This study
modeled the probability that certain antimicrobial
regimens recommended or indicated for use in the
treatment of secondary peritonitis will achieve micro-
biologic success against aerobic bacteria implicated
therein. The study incorporates the pharmacokinetic
and pharmacodynamic properties of each antibiotic,
along with recent resistance data from 2 global sur-
veillance studies. Pharmacodynamic profiling steps
into the gap between available clinical information
and known resistance developments whose impact
may not yet be realized. Furthermore, it has been re-
ported to be predictive of clinical success.3°-33 Entero-
cocci were excluded from the primary analysis of this
study in order to provide a better comparison of an-
tibiotic agents; however, they were included in a par-
allel analysis. The primary analysis predicted overall
CFRs _>90% for the following agents: cefepime 1 and
2 g q12h and ceftazidime 2 g qSh administered as
combination therapy with metronidazole, and imipen-
em 500 mg q6h and piperacillin/tazobactam 3.375 g
q6h as monotherapeutic regimens. Ceftriaxone and le-
vofloxacin achieved predicted CFRs <80%. Predicted
CFRs for tigecycline 50 mg q12h were between
84.5% and 87.6%, depending on whether the analy-
sis included Enterococcus spp.
The methods used in this study were designed to re-
veal the relative strengths and weaknesses of these
agents by modeling their performance against each in-
dividual species, and then weighting the results by the
prevalence of the species in cIAIs to determine the
agent's overall probability for achieving microbiolog-
ic success against an unknown organism drawn from
the overall bacterial population. The 4 antibiotic
agents that were determined to have a predicted over-
all CFR _>90% in the primary analysis would be desir-
able choices for empiric therapy because their success
was attributable to reliable coverage of pathogens that
have been found to be highly prevalent in secondary
893
Clinical Therapeutics

°~
E c'-I .~- CO
O
r~ Oh
n~
Z
~E
°~
0 "~

C~_ Oh I'-.. ~,~ cO O I'-.. Oh C'.I I.~

°~ 04 ch
Q_ Oh Oh Oh Oh Oh Oh Oh ~O Oh
E
O

°~

o cO '..o '..o '..o co

~ ~ ch ~ ch L~
b'l
Oh Oh '..O ~ '..O '..O oh cO I~.

u') O I~. ~_^

O
ch o
q~ ~ ~ Oh Oh Oh Oh or) Oh Oh Oh
Q_ O
._~ ~q

°~ D._
n~ m
m
n~
r./ .~ c'-I I~. O I~. CO U')
U ~ ~ ~" o
cO ~O
E 0
"E0 E-6
-t3 c
L~
O
0 -C~ s_
O

r-. ~o

C1_ "-~

~.~_
I.I_

o~ -~ ~ ._~
0

0
~o ~ ~ ~- .~
0
Oh Oh Oh Oh Oh Oh Oh Oh Oh

bn ._~ -c~

~,~ u~ O ~
E
Oh Oh ~ Oh ~ Oh Oh I~ Oh Oh "- "~c~_

o-~ ~
° lO
'd ..c
_~o~._~
°~
L~ E
•~ E "~- C)- "~- C)-

894 Volume 29 Number 5

 K.J. Eagye et al.

Table III. Cumulative fraction of response (CFR) of antibiotic regimens against pathogens implicated in compli-
cated intra-abdominal infections, by weighted prevalence, including and excluding Enterococcus spp.

Excluding Enterococcus spp Including Enterococcus spp

95% 95% 95% 95%

Confidence Confidence Confidence Confidence
Lower U pper Lower Upper
Antibiotic Regimen CFR, % Limit Limit CFR, % Limit Limit

Cefepime 1 g ql 2h 92.4 91.6 93.1 79.5 78.3 80.6

Cefepime 2 g ql 2h 95.3 94.6 95.8 82.5 81.4 83.5
Ceffazidime 1 g q8h 86.5 85.5 87.4 74.6 73.4 75.8
Ceffazidime 2 g q8h 94.2 93.5 94.8 81.9 80.8 82.9
Ceffriaxone 1 g q24h 75.7 74.5 76.9 64.8 63.4 66.1
Ceffriaxone 2 g q24h 78.2 77.1 79.4 67.0 65.7 68.3
Imipenem 500 mg q6h ~ 96.8 96.3 97.3 93.4 92.7 94.1
Levofloxacin 750 mg q24h 76.3 75.1 77.4 66.4 65.1 67.7
Piperacillin/tazobactam
3.375 g q6h ~ 91.2 90.4 92.0 88.4 87.5 89.2
Tigecycline 50 mg ql 2h ~ 84.5 83.4 85.4 86.7 85.8 87.6

*Denotes monotherapeutic choices; all other agents require concomitant administration ofmetronidazole for coverage of anaer-
obic organisms.

peritonitis. When considering the suitability of the
other agents for empiric therapy, examination of their
performance against individual species may be help-
ful. Since tigecycline 50 mg q12h appeared to reliably
cover highly prevalent organisms, including E coli,
Streptococcus and Enterococcus spp (98.5%, 99.6%,
and 99.6% CFR, respectively), but not P aeruginosa
(1.9% CFR), it may be clinically useful in cases in
which P aeruginosa is less likely to be encountered (ie,
community-acquired disease). Levofloxacin achieved
its bactericidal threshold against E coli in <75% of the
trials. Coverage of E coli is particularly important be-
cause this species constitutes a meaningful portion of
the organisms associated with secondary peritonitis.
In the present study, based on data derived from a
double-blind, randomized, comparative Phase III
clinical trial of ertapenem and piperacillin/tazobactam
for the treatment of clAIs in 633 patients (intent-to-
treat population), 21 E coli accounts for 48% of all
aerobic bacteria encountered. Two European investi-
gations of intra-abdominal infections, a retrospective
chart review and prospective study, reported E coli as
the most frequently encountered pathogen in clAIs,
with prevalence rates of 40% and 47%, respectivly. 7,s
Thus, reliable coverage of E coli should be an impor-
tant consideration in the selection of appropriate em-
piric therapy for patients with secondary peritonitis.
Although some investigations suggest that targeting
enterococci does not improve clinical outcomes in
cIAIs, polymicrobial infections involving Entero-
coccus spp have been associated with increased mor-
tality (21% vs 4% in postoperative cIAI patients
where Enterococcus was not isolated). 34-36 Imipenem,
piperacillin/tazobactam, and tigecycline have been ob-
served to exhibit activity against Enterococcus fae-
calis. However, the database used in this study did not
distinguish between E faecalis and Enterococcus fae-
cium--against which imipenem has been reported to
be inactive. 37 In the parallel analysis, which included
Enterococcus spp, only imipenem achieved a predict-
ed CFR >90%. Piperacillin/tazobactam and tigecy-
cline appeared to be comparable overall, with CFRs
of 88% and 87%, respectively. Tigecycline had the
highest predicted CFR (99.6%) against Enterococcus
spp, followed by imipenem and piperacillin/tazobactam,
with predicted CFRs of 74.2% and 72.4%, respectively.

May2007 895
 Clinical Therapeutics
The results of the primary analysis were similar to
those of previous pharmacodynamic analyses of an-
tibiotic regimens for the treatment of secondary peri-
tonitis by the present investigators. 15,16In those studies,
carbapenems and ~3-1actam/~3-1actamasecombinations
appeared to provide reliable coverage of the modeled
pathogens, and cephalosporins also appeared to be
adequate choices. The fluoroquinolone regimens ap-
peared to exhibit the same lack of coverage against
E coli as that predicted for these agents in the present
study. Since the modeling techniques used in the pres-
ent analysis were more sophisticated than those of the
previous study, and a broader survey of organisms
was assessed, the present findings are considered to
support those of the earlier investigations by the pres-
ent authors.
Because tigecycline was not investigated in the pre-
vious studies, data are not available for comparison.
However, an analysis of pooled efficacy and safety
data from clinical trials reported comparable clinical
cure rates for tigecycline and imipenem among the mi-
crobiologically evaluable group of patients (86.1%
and 86.2%, respectively).3s These findings differ from
those of the present study, which indicate a statisti-
cally significant--although not entirely dissimilar--
difference between these 2 drugs, illustrating the
common conundrum of statistical versus clinical sig-
nificance. Although ex-vivo analyses have been re-
ported to be predictive of clinical results for
comparison of antibiotics within the same analysis,
comparing the results of specific trials can be difficult
because they may be based on different assumptions.
In the present analysis, microbiologic response was
isolated from other factors that could potentially in-
fluence clinical response so that distinctions could be
made solely on the particular characteristics of each
agent. The specific assumptions of this model were
that appropriate surgical source control could always
be obtained and that each antibiotic/pathogen combi-
nation had an equal chance of success (since all other
factors were held constant). This may not be the case
in the clinical setting, however, particularly with
polymicrobial infections. The model used in the pres-
ent study does not predict efficacy against polymi-
crobial infections specifically, but rather to each or-
ganism individually that might be present in
clAIs. In actual clinical practice, the eradication of
one pathogen may affect the growth or eradication of
another.
896
In terms of microbiology, the surveillance data used
in this analysis may reflect the activity of isolates that
are different from, and probably more resistant to
treatment than, those encountered in the referenced
trial, 38 in part, because of the practice of discontinu-
ing patients whose cultures revealed organisms resis-
tant to either study drug. Furthermore, the prevalence
of species in the clinical trial may be different from
that modeled in this study. While the clinical trial did
not describe the prevalence or MIC distribution of the
P aeruginosa isolates encountered, levels lower than
those found in the present analysis could explain the
comparable clinical cure rates found for imipenem
and tigecycline. Indeed, the clinical trial data did re-
port equal eradication rates against P aeruginosa,
where the present study predicted a very large differ-
ence in CFR for the 2 compounds (76.7% for imipen-
em, but only 1.9% for tigecycline). Similarly, the trial
found comparable eradication rates for the 2 drugs
versus E faecalis, although no data regarding E faeci-
urn or VRE specifically was shown. However, the trial
determined eradication by clinical response, noting
that follow-up cultures were not obtained for a majori-
ty of patients, and the CIs around these results are
wide. (For example, for eradication of E faecalis, the
trial reported a rate of 78.8% for tigecycline [95% CI,
61.1%-91.0%].)
Research methods used to predict real-world sce-
narios often rely on assumptions that may or may not
hold true in actual clinical practice. The primary as-
sumption of this analysis was that the MIC distribu-
tions that an agent will encounter when used as
empiric therapy in a secondary peritonitis patient will
resemble the MIC distributions found worldwide,
which are modeled in this study. Clearly, local--and
institutional--MIC distributions may differ. However,
the susceptibility profiles derived from the TEST and
SENTRY databases were similar to those found using
either global or North American surveillance data in
previous studies by the present investigators. 39 Be-
cause the resistance profiles of target isolates strongly
influence microbiologic success, local data should be
incorporated into empiric therapy decisions whenever
possible. A secondary assumption of this analysis was
that the prevalence of the modeled species (derived
from a 2003 clinical trial) fairly reflects current local
prevalence. Assumptions external to the analysis in-
clude those elements whose absence would render
even the most efficacious antibiotic moot, such as
Volume 29 Number 5

the successful attainment of source control and a
sufficiently immunocompetent host. Factors that are
unrelated to the eradication of organisms but could
potentially affect clinical outcomes, such as underly-
ing disease states and adverse-event rates, were as-
sumed to be equal for the entire simulated population.
Clinical trials and ex-vivo analyses, such as the pres-
ent study, are limited representations of the genuine
clinical environment. Each method offers a different
contribution to research, and results should be consid-
ered in the light of the compromises inherent in the de-
sign and execution of each type of study. Pharma-
codynamic profiling may be useful for revealing the
strengths and weaknesses of antimicrobial agents that
may be less apparent when they are assessed using
broader measures. Investigations that incorporate these
techniques with the most recent available bacterial re-
sistance data can provide valuable information for
making distinctions between antimicrobial compounds
that otherwise appear to be equivalent.
CONCLUSIONS
A variety of factors should influence the choice of em-
piric therapy when selecting antimicrobials for sec-
ondary peritonitis, including pharmacokinetic and
pharmacodynamic properties of the drug, and emerg-
ing resistance profiles. Local antibiograms should be
consulted when possible. Accounting for these factors,
combination therapy with metronidazole and either
cefepime 1 or 2 g q12h or ceftazidime 2 g qSh was
predicted to provide an enhanced probability of mi-
crobiological success, as was m o n o t h e r a p y with
imipenem 500 mg q6h or piperacillin-tazobactam
3.375 g q6h. Tigecycline has been found to be equiv-
alent to imipenem in a recent clinical trial of intra-
abdominal infections; our results suggest that these
compounds show their best effect versus different im-
plicated species. If VRE is a concern, or if enterococ-
cal coverage is desired but pseudomonal coverage is to
be avoided (so as not to apply selective pressure), then
tigecycline 50 mg q12h would make an attractive new
choice. As reported earlier, combination therapy
based on levofloxacin or on ceftriaxone may be less
desirable due to emerging resistance in E coli or in
other gram-negative aerobic bacteria.
A C K N O W L E D G M ENTS
The authors thank Ron Jones, MD, and members of
JMI Inc., North Liberty, Iowa, for supplying MIC
May2007
K.J. Eagye et al.
data from the SENTRY Program database. The
authors also thank Wyeth Pharmaceuticals for its con-
tribution of MIC data from the TEST surveillance
program.
REFERENCES
1. Barie PS. Serious intra-abdominal infections. Curt Opin
Crit Care. 2001 ;7:263-267.
2. McClean KL, Sheehan GJ, Harding GK. Intraabdominal
infection: A review. Clin Infed: Dis. 1994;19:100-116.
3. Marshall JC, Innes M. Intensive care unit management of
intra-abdominal infection. Crit Care Med. 2003;31:2228-
2237.
4. Onderdonk AB, Cisneros RL, Finberg R, et al. Animal
model system for studying virulence of and host response
to Bacteroides firagilis. Rev Infect Dis. 1990;12(Suppl 2):$169-
$177.
5. Montravers P, Gauzit R, Muller C, et al. Emergence of an-
tibiotic-resistant bacteria in cases of peritonitis after in-
traabdominal surgery affects the efficacy of empirical
antimicrobial therapy. Clin Infect Dis. 1996;23:486-494.
6. Falagas ME, Barefoot L, Griffith J, et al. Risk factors lead-
ing to clinical failure in the treatment of intra-abdominal
or skin/soft tissue infections. EurJ Clin Microbiol Infect Dis.
1996;15:913-921.
7. Cattan P, Yin DD, Sarfati E, et al. Cost of care for inpa-
tients with community-acquired intra-abdominal infec-
tions. EurJ Clin Microbiol Infect Dis. 2002;21:787-793.
8. Krobot K, Yin D, Zhang Q, et al. Effect of inappropriate
initial empiric antibiotic therapy on outcome of patients
with community-acquired intra-abdominal infections re-
quiring surgery. Eur J Clin Microbiol Infect Dis. 2004;23:
682-687.
9. Sturkenboom MC, Goettsch WG, Picelli G, et al. In-
appropriate initial treatment of secondary intra-abdomi-
nal infections leads to increased risk of clinical failure and
costs. BrJ Clin Pharmacol. 2005;60:438-443.
10. Solomkin JS, Mazuski JE, Baron EJ, et al, for the Infectious
Diseases Society of America. Guidelines for the selection
of anti-infective agents for complicated intra-abdominal
infections. Clin Infect Dis. 2003;37:997-1005.
11. Wong PF, Gilliam AD, Kumar S, et al. Antibiotic regimens
for secondary peritonitis of gastrointestinal origin in
adults. Cochrane Database Syst Rev. 2005;2:CD004539.
12. Zhanel GG, Homenuik K, Nichol K, et al. The glycylcy-
clines: A comparative review with the tetracyclines. Drugs.
2004;64:63-88.
13. Guay DR. Oritavancin and tigecycline: Investigational an-
timicrobials for multidrug-resistant bacteria. Pkarmaco-
therapy. 2004;24:58-68.
14. Craig WA. Pharmacokinetic/pharmacodynamic parame-
ters: Rationale for antibacterial dosing of mice and men.
Clin Infect Dis. 1998;26:1-10, quiz 11-12.
897
 Clinical Therapeutics
15. Kotapati S, Kuti JL, Nicolau DP.
Pharmacodynamic modeling of beta-
lactam antibiotics for the empiric
treatment of secondary peritonitis:
A report from the OPTAMA pro-
gram. Surf Infect (Larchmt). 2005;6:
297-304.
16. Eagye KJ, Kuti JL, Nicolau DP.
Evaluating empiric treatment op-
tions for secondary peritonitis using
pharmacodynamic profiling. Surf
Infect (Larchmt). 2007;8:215-226.
17. Hoban DJ, Bouchillon SK, Johnson
BM, et al. In vitro activity oftigecy-
cline against 6792 gram-negative
and gram-positive clinical isolates
from the global Tigecycline Evalu-
ation and Surveillance Trial (TEST
Program, 2004). Diagn Microbiol
Infect Dis. 2005;52:215-227.
18. Fedler KA, Biedenbach DJ, Jones
RN. Assessment of pathogen fre-
quency and resistance patterns
among pediatric patient isolates:
Report from the 2004 SENTRY
Antimicrobial Surveillance Program
on 3 continents. Diagn Microbiol Infect
Dis. 2006;56:427-436.
19. Van Wart SA, Owen JS, Ludwig EA,
et al. Population pharmacokinetics
of tigecycline in patients with com-
plicated intra-abdominal or skin
and skin structure infections. Anti-
microb Agents Chemother. 2006;50:
3701-3707.
20. DeRyke CA, Kuti JL, Nicolau DP.
Pharmacodynamic target attain-
ment of six beta-lactams and two
fluoroquinolones against Pseudo-
monas aerugmosa, Acinetobacter bau-
mannii, Escherichia coil, and Klebsiella
species collected from United States
intensive care units in 2004. Phar-
macotherapy. 2007;27:333-342.
21. Solomkin JS, Yellin AE, Rotstein
OD, et al, for the Protocol 017
Study Group. Ertapenem versus
piperacillin/tazobactam in the treat-
ment of complicated intraabdomi-
nal infections: Results of a
double-blind, randomized compar-
ative phase III trial. Ann Surf. 2003;
237:235-245.
898
22. MoutonJW, DudleyMN, CarsO, et
al. Standardization of pharmacoki-
netic/pharmacodynamic (PK/PD)
terminology for anti-infective drugs:
An update. J Antimicrob Chemother.
2005;55:601-607.
23. Turnidge JD. The pharmacodynam-
ics of beta-lactams. Clin Infect Dis.
1998;27:10-22.
24. Walker R, Andes D, Conklin J, et al.
Pharmacodynamic activities of
meropenem in an animal infection
model. In: Program and Abstracts
of the 34th Interscience Conference
on Antimicrobial Agents and
Chemotherapy (Orlando). Wash-
ington, DC: American Society for
Microbiology; 1994:147. Abstract.
25. Wright DH, Brown GH, Peterson
ML, Rotschafer JC. Application of
fluoroquinolone pharmacodynam-
ics. J Antimicrob Chemother. 2000;46:
669-683.
26. Craig WA, Andes DR, Ambrose PG,
et al. In vivo pharmacodynamic ac-
tivity of gatifloxacin, levofloxacin,
and moxifloxacin against Staphylo-
coccus aureus with reduced suscepti-
bilityto fluoroquinolones. In: Program
and Abstracts of the 45th Inter-
science Conference on Antimi-
crobial Agents and Chemotherapy.
Washington, DC: American Society
for Microbiology; 2005.
27. Meagher AK, Ambrose PG, Grasela
TH, Ellis-Grosse EJ. Pharmacokinetic/
pharmacodynamic profile for
tigecycline-a new glycylcycline an-
timicrobial agent. Diagn Microbiol
Infect Dis. 2005;52:165-171.
28. Jones RN, Craig WA, Ambrose PG,
et al. Reevaluation of Enterobac-
teriaceae MIC/disk diffusion zone di-
ameter regression scattergrams for
9 beta-lactams: Adjustments of
breakpoints for strains producing
extended spectrum beta-lactamases.
Diagn Microbiol Infect Dis. 2005;52:
235-246.
29. Newcombe RG. Two-sided confi-
dence intervals for the single propor-
tion: Comparison of seven methods.
Stat Med. 1998;17:857-872.
30. Drusano GL, Preston SL, Hardalo C,
et al. Use of preclinical data for se-
lection of a phase II/111 dose for
evernimicin and identification of a
preclinical MIC breakpoint. Anti-
microb Agents Chemother. 2001;45:
13-22.
31. Bradley JS, Dudley MN, Drusano
GL. Predicting efficacy of antiinfec-
tives with pharmacodynamics and
Monte Carlo simulation. Pediatr
Infect Dis J. 2003;22:982-992, quiz
993-995.
32. Kuti JL, Nightingale CH, Nicolau
DP. Optimizing pharmacodynamic
target attainment using the MYSTIC
antibiogram: Data collected in
North America in 2002. Antimicrob
Agents Chemother. 2004;48:2464-
2470.
33. Kuti JL, Ong C, Lo M, et al.
Comparison of probability of target
attainment calculated by Monte
Carlo simulation with meropenem
clinical and microbiological re-
sponse for the treatment of compli-
cated skin and skin structure infec-
tions. Int J Antimicrob Agents. 2006;
28:62-68.
34. Barie PS, Christou NV, Dellinger EP,
et al. Pathogenicity of the entero-
coccus in surgical infections. Ann
Surf. 1990;21 2:155-159.
35. Rohrborn A, Wacha H, Schoffel U,
et al. Coverage of enterococci in
community acquired secondary
peritonitis: Results of a randomized
trial. Surf Infect (Larchmt). 2000;1
95-107.
36. Sitges-Serra A, Lopez M J, Girvent M,
et al. Postoperative enterococcal in-
fection after treatment of compli-
cated intra-abdominal sepsis. BrJ
Surf. 2002;89:361-367.
37. Primaxin [package insert]. White-
house Station, NJ: Merck & Co, Inc;
2001.
38. Babinchak T, Ellis-Grosse E, Dartois
N, et al, for the Tigecycline 301
Study Group and the Tigecycline
306 Study Group. The efficacy and
safety oftigecycline for the treatment
of complicated intra-abdominal in-
Volume 29 Number 5
 K.J. Eagye et al.

fections: Analysis of-pooled clinical

trial data. C[in Infect Dis.
2005;41(Suppl 5):$354-$367.
39. Kuti JL, Nicolau DP. Making the
most of-surveillance studies: Sum-
mary of-the OPTAMA Program.
Diag,n Microbio[ Infect Dis. 2005;53:
281-287.

Address correspondence to: David P. Nicolau, PharmD, FCCP, Center for

Anti-Infective Research and Development, Hartford Hospital, 80 Seymour
Street, Hartford, CT 06102. E-mail: dnicola@harthosp.org

May2007 899