Pseudomonas aeruginosa Ventilator-Associated Pneumonia: Comparison of Episodes Due to Piperacillin-Resistant versus Piperacillin-Susceptible Organisms
J. L. Trouillet
Service de Reanimation Medicale
Laboratoire de Bacteriologie
Hopital Bichat Assitance Publique-Hopitaux de Paris
Received 19 July 2001; revised 26 November 2001; electronically published 15 March 2002. 47-83 Boulevard de l'Hopital, 75651 Paris Cedex 13,
We sought to determine the epidemiological characteristics of patients in an intensive care unit (ICU) who developed ventilator-associated pneumonia (VAP) caused by piperacillin-resistant Pseudomonas aeruginosa (PRPA; n p 34) or piperacillin-susceptible P. aeruginosa (PSPA; n p 101). According to univariate analysis, the factors associated with the development of PRPA VAP were presence of an underlying fatal medical condition, immunocompromised status, longer previous hospital stay, less-severe illness at the time of ICU admission, duration of mechanical ventilation before onset of VAP, number of classes of antibiotic received, and previous exposure to imipenem or fluoroquinolone. Multivariate logistic regression analysis identified the following significant independent factors: presence of an underlying fatal medical condition (odds ratio [OR], 5.6), previous fluoroquinolone use (OR, 4.6), and initial disease severity (OR, 0.8). We concluded that the clinical characteristics of patients who develop PRPA VAP differ from those of patients who develop PSPA VAP. Restricted fluoroquinolone use is the sole independent risk factor for PRPA VAP that is open to medical intervention.
Pseudomonas aeruginosa, one of the main pathogenic
agents responsible for nosocomial pneumonia
(especially among patients who have undergone intubation
and have been hospitalized in intensive care units
[ICUs]), ranks second among all pathogens reported
to the National Nosocomial Infection Surveillance
System or evidenced in the European Prevalence of
Infections in Intensive Care Study [1, 2].
Ventilatorassociated pneumonia (VAP) caused by this
microorganism remains a severe and dreaded complication
. Compared with community-acquired strains,
nosocomially acquired P. aeruginosa isolates tend to
be more resistant (often, to multiple classes of
antibiotics). Thus, infection with multidrug-resistant P.
aeruginosa has been a growing problem in medical
facilities, and the occurrence of infections due to
strains that are resistant to almost all commercially
available antibacterial drugs has become a
not-uncommon event .
Although a substantial proportion of cases of
nosocomial infection are caused by multidrug-resistant P.
aeruginosa, there have been few published studies of
this topic, with the exception of studies of patients with
cystic fibrosis. The few published investigations usually
have involved small numbers of patients with mostly
extrapulmonary infections, have lacked adequate
control subjects, or both [8, 9]. The aim of the present
study was to compare the epidemiological
characteristics, risk factors, and outcomes of patients who received
mechanical ventilation (MV) and developed antipseudomonal
penicillinresistant P. aeruginosa (PRPA) VAP with those who
developed penicillin-susceptible P. aeruginosa (PSPA) VAP.
Historically, carbenicillin was the first penicillin found to be active
against P. aeruginosa, but, because piperacillin has become the
reference drug, the piperacillin susceptibility of the P. aeruginosa
strain is used as a marker of multidrug resistance.
PATIENTS, MATERIALS, AND METHODS
Study population. The study was conducted from January
1994 through August 1999 in one ICU in a 1200-bed Parisian
university hospital (Ho pital Bichat) that serves as both a referral
center and a first-line treatment facility. The only remarkable
characteristic of this 17-bed ICU is its recruitment of a large
number of patients who experience multiple-organ failure
following cardiac surgery (20% of all patients admitted to the ICU).
All hospitalized patients who received mechanical ventilation
(MV) for 148 h were eligible for the study, if pneumonia was
suggested on the basis of clinical criteria. All the data were
prospectively collected and stored in a specially designed database.
It should be noted that, during the study period, no systemic
antibiotic regimen was prescribed for nosocomial pneumonia
prophylaxis or selective decontamination of the digestive tract.
The antimicrobial prescription policy, which was systematically
applied by medical staff, was to prescribe the
narrowest-spectrum antibiotic as soon as the susceptibility of the causative
bacteria became known. For all patients, a heat-moisture
exchanger was positioned between the Y-piece and the patient
and was changed every 48 h . The oropharyngeal cavity
was carefully cleaned with antiseptic solution (hexamedine) 4
times daily. To prevent nosocomial infection, other measures
were also applied systematically, including effective surveillance,
staff education, identification of high-risk patients, proper
isolation techniques, and such practices as handwashing and
wearing gloves for each patient contact.
Diagnosis of VAP. VAP was defined as any lower
respiratory tract infection that developed after a patient received MV
for 2 days. The criteria for clinical suspicion of pneumonia were
presence of a new or persistent lung opacity on chest radiographs
plus 2 of the following findings: presence of either fever
(temperature, 38.3 C) or hypothermia (temperature, !36 C), a
WBC count of 110,000 cells/mm3 or !5000 cells/mm3, and
purulent endotracheal aspirate. Every patient suspected of having
pneumonia underwent fiberoptic bronchoscopy so that protected
specimen brush (PSB) and bronchoalveolar lavage (BAL) fluid
samples could be obtained. Specimens were collected and
processed according to procedures extensively described elsewhere
[11, 12]. Nonbronchoscopic BAL was performed only in rare
cases, and its results were taken into account to confirm VAP
. Thus, VAP was diagnosed on the basis of microbiological
findings from analysis of PSB and BAL samples, according to
the following thresholds for significance (i.e., significant
thresholds): PSB culture yielding 103 cfu/mL, 2% of recovered
cells containing intracellular bacteria on direct examination of
BAL fluid samples, and/or BAL culture yielding 104 cfu/mL.
For nonbronchoscopic BAL, only the last 2 criteria were taken
into account. The presence of 2 of the aforementioned criteria
was always required for the diagnosis of VAP. VAP was considered
to be caused by P. aeruginosa when PSB and/or BAL specimens
yielded significant concentrations of this pathogen. If
microorganisms other than P. aeruginosa grew at rates above the
significant threshold, or if 2 different strains of P. aeruginosa grew to
concentrations greater than the significant thresholds, then the
episode was categorized as a polymicrobial VAP episode.
Microbiologic methods. P. aeruginosa was identified by
standard microbiological methods. Piperacillin susceptibility
was determined by the disk-diffusion test. According to the
criteria of the Antibiogram Committee of the French Society
for Microbiology, the organism was considered susceptible
when the inhibition diameter was 18 mm, intermediate
(denoting intermediately susceptible) when the diameter was
1217 mm, and resistant when the diameter was !12 mm,
all for a disk content of 75 mg of antimicrobial agent .
Intermediate susceptibility to piperacillin was considered to be
resistance, because piperacillin was never prescribed for
intermediate strains in such cases. Susceptibilities to ticarcillin,
ticarcillin-clavulanic acid, piperacillin-tazobactam, ceftazidime,
imipenem, amikacin, or ciprofloxacin were systematically
evaluated by use of the disk-diffusion test . For polymicrobial
VAP episodes, bacterial identification and susceptibility tests
were systematically performed, and their results were taken into
account in the choice of antimicrobial therapy.
Risk factors associated with PRPA VAP. Patients with
PRPA VAP were compared with patients with PSPA VAP to
determine the risk factors for developing PRPA VAP. For the
purpose of the study, only the first episode of P. aeruginosa
VAP was taken into account. If 2 different P. aeruginosa strains
(1 piperacillin-susceptible strain and 1 piperacillin-resistant
strain) were isolated during the same episode, and if they grew
to concentrations greater than the previously defined significant
thresholds, the episode was classified as having been caused by
the piperacillin-resistant strain.
For purposes of comparison, the following information on
each patient was collected within the first 24 h after admission
to the ICU: age; sex; severity of the underlying medical
condition (stratified, according to the criteria of McCabe and
Jackson, as rapidly fatal, ultimately fatal, or not fatal );
immunocompromised status (defined according to criteria used
in the determination of the Acute Physiology and Chronic
Health Evaluation [APACHE] II score ). For HIV-positive
Characteristic or finding
At the time of ICU admission
Age, mean years SD
Underlying medical condition (ultimately or rapidly fatal)
Chronic obstructive pulmonary disease
Direct admission to the ICU
Duration of hospitalization before ICU admission, mean days
APACHE II score, mean SD
Reason for MV
Postoperative respiratory failure
ARDS at admission
ODIN score, mean SD
At the time of diagnosis of VAP
Duration of MV before onset of VAP, mean days SD
Temperature, mean C SD
WBCs 103/mL, mean no. SD
Pao2/Fio2, mean mm Hg SD
Radiological score, mean SD
No. (%) of polymicrobial VAP episodes
Associated with multidrug-resistant microorganismsa
No (%) associated with bacteremia
Received previous antibiotic therapy
No. of antibiotic classes received,b mean no. SD
(n p 101)
NOTE. Data are no. (%) of patients, unless indicated otherwise. APACHE, Acute Physiology and Chronic Health
Evaluation; ARDS, acute respiratory distress syndrome; MV, mechanical ventilation; ODIN, organ dysfunction and/
or infection; Pao2/Fio2, partial pressure of oxygen in the alveoli/fraction of inspiried oxygen; PRPA, piperacillin-resistant
P. aeruginosa; PSPA, piperacillin-susceptible P. aeruginosa.
a Methicillin-resistant Staphylococcus aureus, Acinetobacter baumannii, and Stenotrophomonas maltophilia.
b See table 2 for details.
patients, AIDS was defined according to criteria used in the
Simplified Acute Physiology Score (SAPS) II .
Corticosteroid treatment was considered long term if it was 11 month
in duration, and it was considered high-dose treatment if the
dosage was 1 bolus of 5 mg/kg given daily. Also collected
was information on referral from another hospital or direct
admission from our emergency department (i.e., direct ICU
admission); prior duration of hospital stay; indications for
receiving MV (on the basis of the classification described by
Zwillich et al. ); presence of pneumonia at admission;
presence of chronic obstructive pulmonary disease and acute
respiratory distress syndrome ; APACHE II score ; and
organ dysfunction and/or infection (ODIN) score .
The parameters recorded at the time of each episode of P.
aeruginosa VAP were temperature; WBC count; partial pressure
of oxygen in the alveoli (Pao2)/fraction of inspiried oxygen
(Fio2); radiological score (according to the definition of Fagon
et al. ); associated bacteremia; duration of MV before the
diagnosis of VAP (day of the fiberoptic bronchoscopy; recorded
as a continuous variable); presence of other causative
organisms; and use of any antimicrobial agent for 124 h during the
15 days preceding diagnosis of VAP (day of the fiberoptic
bronchoscopy). In addition, we recorded for each patient the use
of any of the following 5 antibiotic classes during the 15 days
preceding the diagnosis of VAP: imipenem, third-generation
cephalosporin, aminoglycoside, fluoroquinolone, and/or other
antibiotics (defined by exclusion).
Outcome. A recurrent episode of P. aeruginosa VAP was
defined by (1) occurrence of an episode at least 72 h after
clinical resolution of a prior episode, (2) presence of the criteria
for clinical suspicion of pneumonia, as defined in the
Diagnosis of VAP subsection above, (3) a quantitative
bronchoscopic specimen culture that was positive for P. aeruginosa, and
(4) absence of evidence of a new extrapulmonary source of
infection. Finally, the duration of MV, measured from the day
of VAP diagnosis, and the mortality rate among patients in the
ICU were recorded.
Statistical analyses. Potential risk factorsthat is, all the
Fluoroquinolone Other No. (%) of patients with PRPA VAP
(n p 34)
(n p 101)
parameters listed abovewere subjected to univariate analysis
(comparison of means) to identify those that had a significant
association with PRPA VAP. Continuous variables were
compared using Students t test; when that was not appropriate,
the Mann-Whitney U test was used. The x 2 test was used for
categorical variables; when they were not appropriate, Fishers
exact test was used. Differences between groups were considered
to be significant for variables for which P ! .05.
Only those variables for which a p 0.05 were entered into
the model of logistic regression analysis to determine
independent risk factors for developing PRPA VAP. The discriminating
ability of the model was assessed by use of the area under the
receiver operating characteristic (ROC) curve. Mortality rates,
duration of MV, and susceptibility of strains to other antibiotics,
according to piperacillin resistance, were also determined.
For 135 (34.4%) of the 393 consecutive patients who developed
1 episode of VAP, cultures of BAL fluid and/or PSB samples
yielded P. aeruginosa at concentrations above the significant
thresholds. PRPA isolates were recovered from 34 patients
(25.2%) and PSPA isolates from 101 patients (74.8%). The
epidemiological characteristics of VAP caused by the 2 types of
Pseudomonas strains are reported in tables 1 and 2.
For 127 patients, diagnosis of Pseudomonas VAP was
established on the basis of culture results for PSB and BAL samples
obtained during fiberoptic bronchoscopy; for the remaining 8
patients, it was established with BAL fluid samples obtained
blindly. In 48 episodes (35.6%), 1 other pathogen was isolated
at a significant concentration. Details regarding the species of
the other pathogens are indicated in table 3. The incidence of
polymicrobial VAP episodes was higher in the PSPA group
(39.6%) than in the PRPA group (23.5%), but the difference
was not statistically significant (P p .10).
Distributions of PRPA and PSPA VAP episodes, according
to year of diagnosis, are shown in figure 1; no significant trend
was observed (P p .37). Moreover, throughout the entire study,
no period in which there was a peak in the number of P.
aeruginosa VAP episodes was observed, no epidemic strain was
identified, and no temporal-spatial aggregation of cases was
Risk Factors for Developing PRPA VAP
Univariate analysis. According to a univariate analysis
comparison of means, the factors that were significantly associated
with PRPA VAP were the presence of an underlying fatal
medical condition (P p .03), immunocompromised status (P p
.004), longer previous stay in the hospital ICU (P p .009),
lower APACHE II score (P ! .0002), lower ODIN score at the
time of admission to the ICU (P p .03), longer duration of
MV at the time of VAP diagnosis (P p .05), more different
classes of antibiotics administered before VAP (P p .01), and
prior antimicrobial therapy with imipenem (P p .04) or a
fluoroquinolone (P p .007). A third-generation cephalosporin
had been prescribed more frequently to the PRPA group, but
the difference between groups did not reach statistical
significance (P p .08).
Multivariate analysis. When all the variables that were
significantly different in univariate analysis were subjected to
logistic regression multivariate analysis, 3 remained in the final
model: presence of an underlying fatal medical condition, prior
use of a fluoroquinolone, and APACHE II score (table 4). The
odds ratio associated with the APACHE II score is expressed
No. of microorganisms
In PRPA In PSPA
NOTE. MRSA, methicillin-resistant Staphylococcus aureus; MSSA,
methicillin-susceptible S. aureus.
a Eight polymicrobial VAP episodes.
b Forty polymicrobial VAP episodes.
c Neisseria species, Haemophilus species, and diphtheroids.
d The total values are greater than values listed in footnotes a and b because
11 microorganism was associated with individual episodes.
Figure 1. Distribution of episodes of Pseudomonas aeruginosa
ventilator-associated pneumonia, according to year of diagnosis. Only
episodes that occurred during the first 8 months of 1999 were included.
PRPA, piperacillin-resistant P. aeruginosa; PSPA, piperacillin-susceptible
per point accorded and indicates that this parameter is a risk
factor when lower values are recorded at admission. The overall
discriminating value of this multiple regression model to
predict the probability of developing PRPA VAP, as assessed by
ROC analysis, was 0.82. Table 5 shows the percentages of PRPA
VAP episodes, according to the presence or absence of a fatal
underlying disease and previous exposure to a fluoroquinolone.
Ten patients (7 in the PSPA group and 3 in the PRPA group)
had recurrence of P. aeruginosa VAP (i.e., relapse with the
previous strain or infection with a new one). The second episode
developed at a median of 17.4 days (range, 636 days) after
the initial episode. For 6 patients, the antibiotic-resistance
patterns of strains isolated during this second episode differed from
those of strains isolated during the first episode.
The mean duration of MV (SD), measured from the day
of diagnosis of VAP, was 17.3 17.4 days for the PRPA group
and 21.0 26.0 days for the PSPA group (P value not
significant). A total of 20 patients (58.8%) in the PRPA group died
versus 50 patients (49.5%) in the PSPA group (P value not
Susceptibility to Other Antibiotics with Antipseudomonal
Activity, According to Piperacillin Resistance
A significant relationship was found between piperacillin
resistance and cross-resistance to other antibiotics active against P.
aeruginosa infections (table 6). When P. aeruginosa strains were
susceptible to piperacillin, they were !20% resistant to other
antibiotics, except for ciprofloxacin. Piperacillin-resistant strains
were 50% resistant to each drug tested. Thirteen strains (9.6%)
were resistant to piperacillin-tazobactam plus ceftazidime plus
imipenem, and 8 strains were resistant to these 3 antibiotics and
amikacin. All strains were susceptible to colistin.
The resistance of strains to imipenem, ceftazidime, or
ciprofloxacin, according to previous administration of imipenem,
a third-generation cephalosporin, or a fluoroquinolone, is
reported in table 7. As expected, patients who had received 1 of
these 3 antibiotics had a higher probability of becoming infected
with a strain resistant to the given antibiotic class. Moreover,
VAP episodes due to imipenem-resistant strains were more
frequently associated with prior fluoroquinolone administration
than were VAP episodes due to imipenem-susceptible strains
(P p .05).
We analyzed 135 consecutive patients who developed an
episode of P. aeruginosa VAP, using strict criteria to define
pneumonia. PRPA VAP episodes (25% of all P. aeruginosa VAP
episodes) were significantly associated with 9 factors. According
to multivariate analysis, a fatal underlying medical condition,
prior use of a fluoroquinolone, and APACHE II score remained
independently associated with PRPA VAP.
The high percentage of VAP episodes due to P. aeruginosa
(34.4%) confirms findings reported elsewhere . This high
percentage can be explained by the fact that the ICU is an area
of endemicity for P. aeruginosa colonization  and, also, by
the characteristics of the studied populations. Most of our
patients were transferred from another ICU or ward after a mean
hospital stay of 12 weeks. All but 1 episode of P. aeruginosa
VAP occurred 17 days after initiation of MV, and all were
classified as late-onset VAP .
We observed a high rate of increased resistance among the
Pseudomonas strains isolated from the patients with VAP seen at
our ICU. This finding has been noted at other European ICUs,
as was shown by Hanberger et al.  in their study comparing
the antibiotic susceptibilities of aerobic gram-negative bacilli in
5 European countries. Except in Sweden, the percentage of Pseu
domonas strains resistant to piperacillin was 114%, reaching 22%
in Portugal and 26% in France , which is similar to our
findings. The percentage of strains resistant to other antibiotics
(e.g., imipenem and ciprofloxacin) was also high (in France, 24%
were resistant to imipenem and 35% were resistant to
Underlying medical condition that is
rapidly or ultimately fatal
Previous exposure to fluoroquinolone
APACHE II score, per point accorded
OR (95% CI)
NOTE. APACHE, acute physiology and chronic health evaluation.
No prior exposure to fluoroquinolone
No fatal underlying medical condition
Fatal underlying medical condition
Prior exposure to fluoroquinolone
No fatal underlying medical condition
Fatal underlying medical condition
no. of patients
floxacin). Similar results have been observed for Pseudomonas
strains in the United States , even if strict comparison of the
studies is made difficult because standards provided by the
National Center for Clinical Laboratory Standards are defined using
a disk content of 100 mg of piperacillin for the same threshold
diameter of 18 mm. Moreover, our data confirm that PRPA
isolates are more likely to be resistant to other antipseudomonal
agents than are PSPA isolates (table 6).
The specific objective of our study was to determine
independent risk factors that lead to the emergence of PRPA VAP.
The results of our multivariate analysis indicate that exposure
to antibiotics appears to be a crucial factor in the emergence
of multidrug-resistant bacteria and confirms previously
reported data for other resistant pathogens [25, 26]. Finally, the
results of the present study highlighted the major role of
fluoroquinolones in the emergence of multidrug-resistant P.
aeruginosa responsible for VAP. These findings are in accordance
with the results of Harris et al. , who found multidrug and
lengthy antipseudomonal antibiotic exposures to be factors that
predispose for the development of an infection with
multidrugresistant P. aeruginosa. One pertinent finding is that recent
exposure to a fluoroquinolone was also found to be an
independent risk factor for carriage of and persistent colonization
with methicillin-resistant Staphylococcus aureus . This
information could provide an extremely simple way to identify
subgroups of patients from whom PRPA are likely to be isolated.
Our multivariate analysis identified 2 other independent risk
for PRPA: less-severe disease at the time of admission to the
ICU, as assessed by APACHE II score, and presence of an
underlying medical condition that is rapidly or ultimately fatal.
It is unclear why the presence of more-severe disease at
admission to the ICU protects patients from infection with
multidrug-resistant P. aeruginosa strains. This finding may indicate
a subgroup of patients who carry this marker, rather than reflect
a direct causal relationship. The major limitations of
interpreting the marker accurately are the observational nature of
the data and the relatively small number of episodes that may
lower the power of this study. The other major predisposing
factor associated with PRPA pneumonia was the presence of
an underlying medical condition, as assessed by the MacCabe
and Jackson score. This factor might reflect a lower degree of
immune function that leads to infections caused by less virulent
strains, as has been claimed for multidrug-resistant strains of
P. aeruginosa .
Two other factors identified in the case-control study by
Arruda et al.  were confirmed by our univariate analysis but
were eliminated in the multivariate analysis:
immunocompromised status and prolonged length of hospital stay before
the occurrence of nosocomial infection. In the case-control
study, the occurrence of nosocomial infections due to P.
aeruginosa resistant to all antimicrobials commercially available in
Spain was associated with the use of immunosuppressants,
neutropenia, or prolonged prior hospitalization. Prolonged length
of hospital stay before infection was also reported by Harris et
al. , but their study did not include any control population.
In our study, immunocompromised patients were few in
number and low in proportion, precluding a powerful analysis of
Percentage of isolates
Imipenem Ciprofloxacin Amikacin PRPA
NOTE. PRPA, piperacillin-resistant P. aeruginosa; PSPA,
piperacillin-susceptible P. aeruginosa.
No. (%) of patients, by previous drug therapy received
Finally, inclusion of polymicrobial VAP episodes could
introduce another limitation in our result, but the rate of
polymicrobial episodes was not statistically different between the
2 groups. Concerning outcome, mortality rates were similar in
the monomicrobial and polymicrobial groups (50.6% versus
54.2%, respectively, without trend, as found by Crouch Brewer
et al. ). In addition, the proportion of associated
multidrugresistant microorganisms, a factor usually associated with a
poorer prognosis, was not statistically different between the
PRPA and PSPA groups (table 1). In summary, the results of
the present study showed that (1) PRPA VAP episodes
frequently occurred in this type of ICU population, (2)
piperacillin-resistant strains were clearly more resistant to other
antipseudomonal drugs than were piperacillin-susceptible strains,
and (3) the mortality rate among patients with PRPA VAP was
Statistical analyses suggested that the clinical characteristics
of patients who subsequently develop PRPA VAP differ from
those of patients with PSPA VAP episodes. Because previous
use of a fluoroquinolone was the only independent factor open
to medical intervention that was identified in the present study,
these data support the recommendation that in-hospital
prescription of broad-spectrum antibiotics, especially
fluoroquinolones, should be restricted.
We thank Nathalie Desruelle for manuscript preparation.