De-escalation therapy in ventilator-associated pneumonia*

Jordi Rello, MD, PhD; Loreto Vidaur, MD; Alberto Sandiumenge, MD; Alejandro Rodríguez, MD;
Belen Gualis, MD; Carmen Boque, MD, PhD; Emili Diaz, MD, PhD

Objective: To evaluate de-escalation of antibiotic therapy in
patients with ventilator-associated pneumonia.
Design: Prospective observational study during a 43-month
period.
Setting: Medical-surgical intensive care unit.
Patients: One hundred and fifteen patients admitted to the
intensive care unit with clinical diagnosis of ventilator-associated
pneumonia. All the episodes of ventilator-associated pneumonia
received initial broad-spectrum coverage followed by reevalua-
tion according to clinical response and microbiology. Quantitative
cultures obtained by bronchoscopic examination or tracheal as-
pirates were used to modify therapy.
Interventions: None.
Measurements and Main Results: One hundred and twenty-one
episodes of ventilator-associated pneumonia were diagnosed.
Change of therapy was documented in 56.2%, including de-
escalation (the most frequent cause) in 31.4% (increasing to 38%
if isolates were sensitive). Overall intensive care unit mortality
rate was 32.2%. Inappropriate antibiotic therapy was identified in
9% of cases and was associated with 14.4% excess intensive care
unit mortality. Quantitative tracheal aspirates and bronchoscopic
samples (58 protected specimen brush and three bronchoalveolar
lavage) were associated with 32.7% and 29.5% intensive care unit
mortality and 29.3% and 34.4% de-escalation rate. De-escalation
was lower (p < .05) in the presence of nonfermenting Gram-
negative bacillus (2.7% vs. 49.3%) and in the presence of late-
onset pneumonia (12.5% vs. 40.7%). When the pathogen remained
unknown, half of the patients died and de-escalation was not
performed.
Conclusion: De-escalation was the most important cause of
antibiotic modification, being more feasible in early-onset pneu-
monia and less frequent in the presence of nonfermenting Gram-
negative bacillus. The impact of quantitative tracheal aspirates or
bronchoscopic techniques was comparable in terms of mortality.
(Crit Care Med 2004; 32:2183–2190)
KEY WORDS: de-escalation; therapy; ventilator-associated pneu-
monia; hospital-acquired pneumonia; imipenem

O ver the 20-yr period in which
animal and clinical investiga-
tions have attempted to iden-
tify the most effective meth-
odology for diagnosing ventilator-
associated pneumonia (VAP) and thus
rapidly implementing therapy, no con-
sensus has been reached on standard care
for either directed or empirical diagnostic
protocols. The debate on the manage-
ment of VAP has often been characterized
as a comparison of directed therapy (i.e.,
guided by microbiological investigations)
vs. empirical therapy, as reflected in the
*See also p. 2344.
From the Critical Care Department, Joan XXIII Uni-
versity Hospital, University Rovira i Virgili, Tarragona,
Spain.
Supported, in part, by CIRIT SGR 2001/414, Dis-
tinció Recerca Universitaria (JR), and RED Respira
(RTIC C03/11).
Presented, in part, at the American Thoracic So-
ciety annual meeting, Orlando, FL, May 2004.
Dr. Rello served on Advisory Boards from Merck,
Pfizer, Astra-Zeneca, and Wyeth.
Copyright © 2004 by the Society of Critical Care
Medicine and Lippincott Williams & Wilkins
DOI: 10.1097/01.CCM.0000145997.10438.28
American Thoracic Society guidelines
(1). Other investigators (2) have used cri-
teria based on different concepts. Cur-
rently, it is agreed that the speed at which
adequate therapy is delivered is a deter-
minant of outcome, but how to reach this
goal remains a controversial question.
Recent clinical research (3– 6) has em-
phasized the importance of avoiding in-
appropriate initial antibiotic therapy and
has suggested that using directed therapy
is problematic. Other reports warn of the
risk of antibiotic overuse (7–10). De-
escalation therapy is a relatively new con-
cept that is currently used in the man-
agement of serious infections (11–14),
particularly serious nosocomial pneumo-
nias. This concept is midway between two
controversial positions: the exclusive use
of an empirical antibiotic prescription af-
ter a clinical diagnosis and the approach
that advocates the use bronchoscopic
techniques to manage patients with VAP.
As the suggested approach shortens anti-
biotic exposure, the selection pressure for
resistance is reduced.
Indeed, streamlining antibiotic ther-
apy in infected patients and narrowing
the antibiotic spectrum of drugs given
empirically according to culture results
are feasible (15, 16). However, the utility
of this approach in VAP has not yet been
evaluated. It appears important to iden-
tify the variables that influence this strat-
egy and to assess how it is executed in a
wide range of hospitals. Moreover, in an
era of evidence-based medicine, little in-
formation is available on the impact of
this strategy on practice. In 1997, our
group reported that de-escalation was ap-
plied in only 6.1% episodes of a VAP co-
hort in which all patients received bron-
choscopic examination and in which the
incidence of Pseudomonas aeruginosa
was near 50% (17). Unfortunately, geo-
graphical variability (18) suggests the
need to compile data on the results of
de-escalation in a range of intensive care
unit (ICUs).
We designed a prospective follow-up
study with the following objectives: a) to
evaluate the practice of de-escalation in a
cohort of patients with VAP; b) to evalu-
ate the impact of microbiological infor-
mation on management of ventilator-
associated pneumonia; and c) to identify

Crit Care Med 2004 Vol. 32, No. 11 2183

the variables that influence de-escalation. Patients with late-onset pneumonia (⬎7 therapy and aims to avoid the overuse of an-
Our hypothesis was that de-escalation days of intubation) or who had chronic ob- tibiotics. The strategy was defined by a two-
was highly dependent on the incidence of structive pulmonary disease were considered stages process. The first stage involves admin-
nonfermentative Gram-negative bacilli at risk of P. aeruginosa and received initial istering broad-spectrum antibiotics. The
identified by microbiological investiga- antipseudomonal combination therapy. At- second stage focuses on simplifying the anti-
tion. tending physicians followed a common proto- biotic therapy. This approach to the treatment
col but were blind to the nature of study, and of VAP involves a) changing the focus from
data were interpreted by two independent in- multiple agents to a single agent if Pseudomo-
MATERIALS AND METHODS
vestigators (LV, AR) to avoid bias in the ther- nas sp. is not present; b) shortening the ther-
During a 43-month period, a prospective apeutic approach. On receiving the results of apy to ⬍5 days if patients presented negative
follow-up study of all intubated patients with the microbiological etiology of VAP, attending cultures and ⬎48 hrs of defervescence; and c)
suspected pneumonia was conducted. In- physicians de-escalated the antibiotic therapy changing from a broad to narrow agent based
formed consent and approval by the Ethics if the microbiological findings and clinical re- on culture data. In brief, patients receiving
Committee were waived due to the observa- sponse permitted. carbapenems were de-escalated to piperacillin-
tional nature of the study. Our protocol in- Definitions. De-escalation implements the tazobactam, and patients receiving piperacil-
cluded microbiological information using initial broad-spectrum empirical antibiotic lin-tazobactam were de-escalated to an anti-
quantitative cultures obtained by bronchos-
copy (with protected specimen brushing [PSB]
alone in immunocompetent subjects) or tra-
cheal aspirate (TA) in all intubated patients in
whom pneumonia was suspected. Two blood
cultures were carried out simultaneously in
most patients, as were cultures of pleural fluid
if present. These procedures were performed
before empirical antibiotic therapy was started
or before new antibiotics were prescribed in
patients with prior antibiotic therapy.
Ventilator-associated pneumonia was sus-
pected when persistent, new pulmonary infil-
trates not otherwise explained appeared on
chest radiographs, in conjunction with puru-
lent respiratory secretions. At least one of the
following criteria were also required: fever
ⱖ38°C and leukocytosis ⱖ10,000 mm3. Con-
firmation of pneumonia was made by absolute
consensus after clinical rounds at a meeting of
three to six attending physicians in our de-
partment. We excluded episodes in which the Figure 1. Details on the implementation of the protocol of management of ventilator-associated
final diagnosis was rejected or was indetermi- pneumonia (VAP): Initial approach. MRSA, methicillin-resistant Staphylococcus aureus.
nate due to the presence of an alternative
cause or in which no agreement on diagnosis
was reached.
Fiberoptic bronchoscopy with PSB or
bronchoalveolar lavage (BAL; weekdays from 8
am to 5 pm) or TA with quantitative cultures
(weekdays 5 pm to 8 am and weekends) was
performed within 6 hrs of the report of the
development of new pulmonary infiltrate. De-
tailed information regarding these diagnostic
procedures has been presented elsewhere (17,
19). No patients received new antibiotics in
the short period between VAP onset and mi-
crobiological testing. The bacterial etiology of
VAP was confirmed if the PSB yielded ⱖ1000
colony-forming units (CFU)/mL, BAL yielded
ⱖ10,000 CFU/mL, or TA yielded ⱖ100,000
CFU/mL. Positive qualitative cultures from
pleural fluid or blood samples also confirmed
the etiology.
The empirical antibiotic choices used to
treat VAP were based on a cycling/mixing ro-
tational protocol, following a recently pub-
lished management algorithm (20, 21). Details
on the implementation of the protocol are Figure 2. Details on the implementation of the protocol of management of ventilator-associated
summarized in Figures 1 and 2. pneumonia: 48-hr reassessment. ATB, antibiotic; ARDS, acute respiratory distress syndrome.

2184 Crit Care Med 2004 Vol. 32, No. 11

pseudomonal cephalosporin in presence of P.
aeruginosa, if possible. In absence of P.
aeruginosa, patients with combination ther-
apy were switched to monotherapy after with-
holding of ciprofloxacin or amikacyn. Simi-
larly, the other agent was changed to a
nonantipseudomonal betalactam.
Appropriate therapy was defined as the use
of at least one antibiotic to which all isolates
were susceptible in vitro from the moment
bronchoscopy was performed. In presence of
P. aeruginosa, at least two active agents (com-
bination therapy) were required.
Clinical resolution was defined in patients
who had complete resolution of all signs and
symptoms of pneumonia in conjunction with
improvement, or lack of progression, of all
abnormalities on the chest radiograph. Deaths
were considered related to the pulmonary in-
fection if they occurred before any objective
response to antimicrobial therapy or if the
pulmonary infection was considered a contrib-
uting factor to death in patients with a comor-
bidity (17).
The excess ICU mortality caused by inap-
propriate initial therapy was determined by
subtracting the ICU crude mortality rate when
the patient was already receiving appropriate
empirical therapy from the crude mortality
rate of patients with inadequate therapy in
whom the antibiotic treatment was modified
after the cultures demonstrated the isolation
of a resistant organism. The microbiological
investigation was classified as relevant if a)
combination therapy was started because of
the presence of P. aeruginosa; b) an inappro-
priate initial regimen was replaced; or c) a
previously appropriate treatment was substi-
tuted with a more rational (lower spectrum)
alternative.
Microbiological Management of Samples.
In patients in whom ventilator-associated
pneumonia was suspected, TA or bronchos-
copy was performed. After the protected spec-
imen brush was transected into a sterile vial
containing 1 mL of sterile lactated Ringer’s
solution, the vial was vigorously shaken for at
least 60 secs to suspend all the material from
the brush. Specimens were immediately sent
to the laboratory for quantitative cultures. Ali-
quots of 0.01 mL were taken from the original
suspension and inoculated into blood agar,
MacConkey agar, buffered charcoal yeast ex-
tract agar, and Sabouraud medium. One
0.001-mL aliquot was also inoculated into
chocolate agar medium. Culture plates were
incubated at 37.8°C under adequate aerobic
and anaerobic conditions; all plates except Sa-
bouraud plates were evaluated for growth at
24 and 48 hrs. For the PSB, bacterial counts of
ⱖ1000 CFU/mL were used as the cutoff point
to diagnose pneumonia. Two serial ten-fold
dilutions were then done on the recovered
bronchoalveolar lavage fluid, and 0.01-mL ali-
quots of the original suspension and each di-
lution were placed onto plates in the same way
as for the protected specimen brush sample.
Crit Care Med 2004 Vol. 32, No. 11
All PSB and BAL isolates were identified by
standard laboratory techniques.
Statistical Analysis. Descriptive analysis
was performed. Means were compared using
Student’s t-test and the Mann-Whitney test.
Proportions were compared using the chi-
square test with Yates correction or Fisher’s
exact test when necessary. Confidence inter-
vals for proportions were obtained assuming
the binomial distribution. Fisher’s exact test
for unpaired samples and McNemar’s test for
paired samples were used to determine the
statistical significance of differences. All p val-
ues and confidence intervals are two-sided. All
interval estimates are 95% confidence inter-
vals.
RESULTS
One hundred and twenty-one episodes
of VAP in 115 patients (six relapses) were
identified during the study period. The
median age of the study population was
60 yrs, 88 (73%) were males, median
Acute Physiology and Chronic Health
Evaluation II score was 15, and median
onset of VAP was after 8 days of mechan-
ical ventilation. VAP developed after 5
days of intubation in 62.8% of episodes.
The case-mix was medical in 47% of the
patients, trauma in 38.3%, urgent sur-
gery in 10.4%, and elective surgery in
4.3%. Underlying illnesses of the study
population are summarized in Table 1.
Microbiological identification was
based on quantitative information from
PSB in 58 cases (47.9%), TA in 58 cases
Table 1. Underlying diseases of 115 patients with ventilator-associated pneumonia
Disease
Medical conditions
Stroke
Acute respiratory failure
Cardiopulmonary arrest
Acute myocardial infarction
Cardiac failure
Neuromuscular disease
Coma
Upper gastrointestinal bleeding
Neoplasias
Near drowning
Pancreatitis
Others
Trauma
Multiple trauma
Head injury
Urgent surgery
Abdominal surgery
Vascular surgery
Neurosurgery
Elective surgery
Vascular surgery
Head and neck surgery
Total
DE, de-escalation.
(47.9%), BAL in only three cases (2.4%),
and blood cultures in two cases (1.8%).
No major complications (bleeding, pneu-
mothorax, or persistent hypoxemia) re-
lated to the bronchoscopic procedures
were documented. A causative pathogen
was identified in 111 of 121 episodes of
VAP, 54 by PSB (44.6%), 52 by TA (43%),
three by BAL (2.5%), and two by blood
cultures (1.6%). Only ten episodes (8.3%)
had unknown etiology. Quantitative TA
and bronchoscopic samples allowed iden-
tification of pathogens in 90% and 93% of
episodes, respectively. The crude mortal-
ity rate, measured as intra-ICU mortality,
of the study population was 32.2%, and
there were no differences between the
patients diagnosed by bronchoscopy or
quantitative TA (29.5% and 32.7%, re-
spectively, Table 2).
The most frequently isolated microor-
ganism was Haemophilus influenzae in 32
episodes (23%), followed by oxacillin-
sensitive Staphylococcus aureus in 30
(21.5%), P. aeruginosa in 21 (15%), Strep-
tococcus pneumoniae in 12 (8.6%), and
Acinetobacter baumannii in 12 (8.6%).
Overall, 139 isolates were identified in 111
episodes, 26 of them being polymicrobial
(23%). The causative pathogens and the
influence on de-escalation are shown in
detail in Table 3.
The most frequently prescribed empir-
ical agents were carbapenems (37 epi-
sodes, 30%), followed by piperacillin-
No. (%) DE, No. (%)
54 (47) 18 (33.3)
13 7
9 2
5 1
5 2
3 1
2 2
2 0
2 0
1 1
1 1
8 0
44 (38.3) 17 (38.6)
26 10
18 7
12 (10.4) 2 (16.6)
6 2
5 0
1 0
5 (4.3) 1 (20)
4 1
1 0
115 38
2185
Table 2. Respiratory cultures in 119 episodes of ventilator-associated pneumoniaa
Microbiological Performed, Positive,
Technique No. No.
TA 58 52
Bronchoscopicb 61 57
IAT, inadequate antibiotic therapy; DE, de-escalation; TA, tracheal aspirate.
a
Two episodes had only blood cultures; b3 bronchoalveolar lavage and 58 protected specimen
brush.
Table 3. Microorganisms identified in 121 cases of ventilator-associated pneumonia
Pathogen
None
Haemophilus influenzae
OSSA
Pseudomonas aeruginosa
Enterobacteriaceae
Acinetobacter baumannii
Streptococcus pneumoniae
ORSA
Others
OSSA, oxacillin-sensitive Staphylococcus aureus; ORSA, oxacillin-resistant S. aureus; DE, de-
escalation.
tazobactam (18.2%), ceftazidime (15.7%),
and cefepime (14.9%). The percentage of
inappropriate therapy for these agents was
8.1%, 4.5%, 0%, and 16.7%, respectively.
Ciprofloxacin was prescribed in 17 epi-
sodes, usually as part of a combination reg-
imen (15 of 17). Indeed, a regimen com-
bining two agents was prescribed in 52.8%
of episodes and was associated with higher
de-escalation rates (odds ratio, 3.05; 95%
confidence interval, 1.34 – 6.96). The per-
centage of inappropriate empirical therapy
in combined regimens was 7.8%, and no
differences in ICU mortality were found
compared with monotherapy (32.8% vs.
31.6%, p ⬎ .20). Table 4 shows these data
in detail and provides information on the
rate of de-escalation for the various agents.
Sensitive strains were identified in 100
episodes, and de-escalation was possible
in 38. The overall de-escalation rate was
thus 31.4%. This figure increased to
34.2% when we excluded episodes with
initial inadequate therapy and to 38%
when excluding episodes with unknown
etiology (Table 5). The initial empirical
therapy was maintained in only 53
(43.8%) episodes, 46 of them despite the
isolation of sensitive pathogens.
Thirty-five of these 46 patients main-
tained therapy unchanged due to micro-
biological information. Nine patients had
oxacillin-resistant Staphylococcus aureus
or A. baumannii and did not have other
alternative sensitive agents. Similarly, 13
2186
Overall, de-escalation was imple-
mented in only 2.7% of episodes with
IAT, DE, Unchanged, Death,
potentially multiresistant pathogens (i.e.,
No.a No.a No. No.
nonfermenting Gram-negative bacilli and
5 17 30 19 oxacillin-resistant Staphylococcus au-
6 21 21 18 reus) compared with 49.3% among the
remaining pathogens (p ⬍ .05). In 23
cases, it was because no alternative
agents with narrower spectrum were
available to provide the option of de-
escalation in episodes caused by multire-
sistant pathogens. Eleven episodes were
Episodes associated with inappropriate empirical
(Polymicrobial) DE, No. (%) regimen. The remaining patients had de-
layed clinical resolution.
10 0 De-escalation occurred in 34.4% of ep-
32 (12) 18 (56.2) isodes diagnosed by bronchoscopic pro-
30 (12) 16 (53.3)
21 (1) 1 (4.7) cedures, compared with 29.3% of epi-
22 (11) 6 (27.2) sodes diagnosed by quantitative tracheal
12 (4) 0 aspirates (absolute difference ⫽ 5.1%, p
12 (11) 4 (25) ⫽ nonsignificant). In 11 (9%) episodes,
4 (0) 0
resistant pathogens were identified and
4 (3) 1 (25)
initial inappropriate therapy was modi-
fied within the first 72 hrs after VAP on-
set. Details of the factors associated with
inappropriate therapy are shown in Table
6. An excess mortality of 14.4% (p ⫽
patients with P. aeruginosa did not have nonsignificant) was estimated, compared
narrower spectrum alternatives. Eight with a 31% ICU mortality rate of episodes
additional episodes were caused by pa- with initial appropriate treatment.
tients with Enterobacteriaceae (e.g., Antibiotics were changed in only three
Klebsiella sp) producing extended-spec- of ten episodes (Fig. 3) in which the eti-
trum betalactamase (ESBL), in whom ology was unknown (Table 5). These
carbapenem was maintained as a first-line three changes were due to clinical dete-
option. Five patients received nonantip- rioration, compared with 12 (18.4%) in
seudomonal agents due to a combination of the episodes with a confirmed pathogen.
early-onset episodes, comorbidities, and Half of these ten patients died and one
Gram-negative stain; in these patients, de- was readmitted with a subphrenic abscess
escalation was not possible. Finally, 11 ep- 1 month after discharge. No de-escalation
isodes (10%) might receive de-escalation was performed in this subset of patients
therapy from broader to narrower spec- because the clinical resolution was de-
trum antibiotics. In these 11 cases, empir- layed for ⬎5 days: mean resolutions for
fever (ⱕ 38°C) and PaO2/FIO2 (ⱖ250)
ical therapy was maintained unchanged
were 5.6 ⫾ 5.2 days and 7.9 ⫾ 6.6 days
because defervescence and other signs of
after VAP onset, respectively. The excess
clinical resolution were delayed. Overall,
mortality for patients with unknown eti-
microbiological information was relevant
ology (50%) was estimated to be 19.4%
to changes in antibiotic therapy in 49 (odds ratio, 2.26 95% confidence interval,
episodes of VAP (40.4%). Antibiotics were 0.61– 8.34) when compared with episodes
changed in four other patients due to with microbiological diagnosis (ICU mor-
adverse events (Fig. 3). tality rate ⫽ 30.6%). Finally, the ICU
De-escalation was performed in 40.7% mortality rate of patients with de-
of patients with early-onset VAP (defined escalation therapy was significantly lower
as VAP developed within the first 7 days of than in patients who maintained a fixed
mechanical ventilation) compared with regimen (18.4% vs. 43.4%, p ⬍ .05).
12.5% of patients with late-onset VAP
(odds ratio, 4.81; 95% confidence inter-
DISCUSSION
val, 1.7–13.6). The mean onset of VAP in
the group of patients in which de- In this observational study, changes in
escalation was performed was 6.4 ⫾ 6.7 initial therapy were documented in more
days vs. 11.1 ⫾ 13.6 days in patients with- than half of the cohort with VAP. De-
out de-escalation (p ⬍ .05). escalation was the most frequent change
Crit Care Med 2004 Vol. 32, No. 11
and was possible in 38% of episodes with etiology were included. Despite the re- negative bacillus suggest that the effec-
sensitive strains. These findings are in duction, the figure is still statistically tiveness of this approach varies according
agreement with earlier reports that mi- higher than the 6.1% reported in a co- to local patterns of sensitivity.
crobial investigation in patients with hort of VAP (17) in which the prevalence Our findings suggest that the diagnos-
pneumonia frequently leads to changes in of P. aeruginosa was 48.3%. These data tic strategy followed to obtain relevant
antibiotic therapy (15–17, 22–25). The and the current observation that de- respiratory samples before the initiation
figure fell to 34.2% and 31% when epi- escalation was significantly less frequent of antibiotic therapy in patients with VAP
sodes with resistant strains and unknown in patients with nonfermenting Gram- was determinant to modify the empirical
regimen chosen. ICU mortality was
19.4% higher in patients with unknown
Table 4. Initial empirical therapy and de-escalating rate etiology, and changes in antibiotic pre-
scription were unlikely. De-escalation
Agent No. (%) De-escalation, No. (%) was never possible in patients with neg-
ative cultures. This should be interpreted
Carbapenem
Overall 37 (30) 13 (35) in the context of possible false-negative
Monotherapy 19 (16) 5 (26.3) cultures associated with limitations in
Combination therapy 18 (14) 8 (44.4) the sensitivity of the technique and the
Amikacin time to resolution of the febrile episode
Overall 27 (22.3) 11 (40.7)
Monotherapy 0 (0) 0 (0)
(⬎5 days). A recent report from Singh et
Combination therapy 27 (22.3) 11 (40.7) al. (26) suggested that in critically ill sur-
Piperacillin/tazobactam gical patients with a pulmonary infiltrate
Overall 22 (18.2) 5 (22.7) and clinical pulmonary infection score
Monotherapy 8 (6.6) 2 (25) (CPIS) ⬍6 after 3 days of therapy, the
Combination therapy 14 (11.6) 3 (21.4)
Ceftazidime regimen can be stopped without any det-
Overall 19 (15.7) 11 (58) rimental effect on outcome. Our protocol
Monotherapy 0 (0) 0 (0) was based on an individual reevaluation
Combination therapy 19 (15.7) 11 (58) of patients (predominantly with medical
Cefepime
Overall 18 (14.9) 7 (39)
conditions) after 3 days of therapy (21) to
Monotherapy 8 (6.6) 3 (37.5) rule out alternative causes of infection or
Combination therapy 10 (8.3) 4 (40) poor clinical resolution (Fig. 2). Interest-
Ciprofloxacin ingly, defervescence in all patients with
Overall 17 (14) 5 (29.4) unproven etiology was delayed ⬎5 days
Monotherapy 2 (1.6) 1 (50)
Combination therapy 15 (12.4) 4 (26.6) after VAP onset.
Amoxicillin/clavulanate When quantitative cultures were ob-
Overall 14 (11.6) 0 (0) tained, the impact of tracheal aspirate
Monotherapy 14 (11.6) 0 (0) and bronchoscopic techniques was com-
Combination therapy 0 (0) 0 (0)
Linezolid
parable in terms of mortality, and de-
Overall 10 (8.3) 6 (60) escalation was possible in an additional
Monotherapy 0 (0) 0 (0) (nonsignificant) 5% of episodes if bron-
Combination therapy 10 (8.3) 6 (60) choscopy was performed. Our interpreta-
Clindamicin tion is that the higher specificity of bron-
Overall 5 (4.1) 5 (100)
Monotherapy 0 (0) 0 (0) choscopic samples facilitates the
Combination therapy 5 (4.1) 5 (100) reduction of the antibiotic spectrum, al-
though the effect might have been even

Table 5. Patients with unknown etiology

VAP Onset,
Case No. Etiology APACHE II ATB 1 ATB 2 Days Change of Therapy Outcome

1 Unknown 14 C 2 No A
2 Unknown 17 CTZ CLIN 2 No A
3 Unknown 18 C 3 No D
4 Unknown 11 AMC 3 Poor clinical evolution D
5 Unknown 14 CRO CLIN 3 Poor clinical evolution A
6 Unknown 15 C 4 No D
7 Unknown 19 CTZ LNZ 5 Poor clinical evolution A
8 Unknown 21 CRO 5 No D
9 Unknown 21 C AKN 7 No D
10 Unknown 22 Q PTZ 23 No A (intraabdominal abscess)

APACHE, Acute Physiology and Chronic Health Evaluation; ATB, antibiotic; VAP, ventilador-associated pneumonia; C, carbapenem; CTZ, ceftazidime;
CLIN, clindamycin; D, death; AMC, amoxicillin plus clavulanate; CRO, ceftriaxone; Q, fluoroquinolones; LNZ, linezolide; AKN, amikacin; PTZ, piperacilli/
tazobactam.

Crit Care Med 2004 Vol. 32, No. 11 2187

 D
e-escalation was
the most impor-
tant cause of anti-
biotic modification, being more
feasible in early-onset pneumo-
nia and less frequent in the
presence of nonfermenting
Gram-negative bacillus.

tion and were eventually discharged alive.

Therefore, providing early effective ther-
apy in any patient with a reasonable sus-
picion of VAP to avoid treatment failure
should form part of the optimal manage-
ment of patients with VAP.
It is significant that de-escalation was
Figure 3. Algorithm detailing changes in antibiotic therapy based on microbiological results.
finally not implemented in more than
half of 100 episodes with sensitive micro-
Table 6. Factors associated with 11 episodes with inadequate therapy organism. Indeed, 11 episodes might
have received de-escalation therapy with
IAT MV, days Microorganism TA/PSB Outcome a narrower spectrum based on the micro-
biological findings alone. In some cases,
1 28 Acinetobacter baumannii PSB Resolution this was due to delays in providing iden-
2 17 Pseudomonas aeruginosa TA Resolution
3 16 Proteus species PSB Resolution tification of the microorganism and sen-
4 15 A. baumannii BAL Death sitivity in the laboratory, whereas in
5 10 A. baumannii TA Death many others it was due to nonresponse to
6 9 ORSA PSB Death empirical therapy. Moreover, 15 patients
7 3 ORSA PSB Resolution
8 3 P. aeruginosa PSB Resolution
required a change of therapy because
9 3 OSSA PSB Death they were classified as treatment failures.
10 2 Klebsiella pneumoniae TA Resolution This observation is consistent with a re-
11 1 P. aeruginosa TA Death cent study (35) in which nonresponse was
found in a surprisingly high 62% of epi-
IAT, inadequate antibiotic treatment; MV, mechanical ventilation; TA, tracheal aspirate; PSB,
sodes. Carbapenems remained the only
protected specimen brushing; BAL, bronchoalveolar lavage; BAS, boric acid solution; ORSA, oxacillin-
resistant Staphylococcus aureus; OSSA, oxacillin-sensitive S. aureus.
active agent in all episodes caused by A.
baumannii as well as many episodes
caused by P. aeruginosa. Carbapenems
were maintained as first-line therapy in
less marked if higher thresholds of quan- protocol (20, 21) requires avoiding the those episodes in which ESBLs producing
titative cultures for TA were required choice of an antibiotic to which the pa- Enterobacteriaceas were identified, based
(27). A policy of shortening therapy (28) tient has been exposed (10). The admin- on reports in which outcomes were sig-
and decreasing the number of antibiotics istration of broad-spectrum antibiotics nificantly different when treated with
given to ICU patients (8, 29) in order to with anti-pseudomonal coverage in other agents (e.g., cephalosporins) (7,
contain the emergence and dissemina- 83.5% of the regimens (Table 5) and the 36).
tion of nosocomial pathogens (13) should familiarity with dominant local patho- Our study has several limitations.
be a priority in the fight against the gens and antibiotic sensitivities (18, 34) First, it is an observational cohort study,
emergence and rapid dissemination of may have contributed to this low rate of not a blind comparative one. Indeed, the
multiresistant bacteria. inappropriate initial antibiotic choice. In study lacks a control group to show the
Effective antibiotic therapy initiated at the current study, we estimated that in- magnitude of changes in antibiotic use
an early stage of infection reduces mor- appropriate early antibiotic therapy de- resulting from the implementation of de-
tality (3, 17, 30 –33). The rate of inappro- spite a microbiologically guided change escalation therapy. However, due to the
priate antibiotic therapy (9%) recorded was associated with an excess mortality of complex course of ICU patients and the
here was lower than the rates reported in 14.4%. Nonetheless, more than half of difficulty of blinding in this setting, it is
previous studies (17). Our patient-based these patients presented clinical resolu- conceivable that no definitive trial will

2188 Crit Care Med 2004 Vol. 32, No. 11

ever be conducted. Second, interpreta-
tion of these studies is complicated by the
heterogeneity of the population at risk,
and the findings may be influenced by
institution-specific variables. Third, defi-
nition of early-onset pneumonia based on
length of hospital stay rather than length
of intubation might provide different
findings. Finally, episodes of early-onset
pneumonia differ from both late-onset
pneumonia and recurrences. They are
usually associated with different patho-
gens, and the impact of different antimi-
crobials may vary from pathogen to
pathogen. Therefore, although the im-
pact on mortality was similar, our study
cannot rule out the possibility that differ-
ent antimicrobial choices may have dif-
ferent impacts on clinical resolution.
Further studies should evaluate the im-
pact on resolution using surrogates of
outcome other than mortality.
We do not know how far our findings
can be generalized to other sites, and
further studies are needed to evaluate de-
escalation strategies in patients with
other case-mixes, severity, epidemiology,
microbial sensitivities, and microbiologi-
cal techniques. As shown in Table 2, most
of our bronchoscopic techniques were
PSB. Similarly, qualitative tracheal aspi-
rates may be associated with different
outcomes. Furthermore, the fact that pa-
tients with early-onset pneumonia could
be treated empirically with less broad
spectrum drugs may have decreased the
rate of de-escalation. Our findings of de-
escalation in VAP cannot be extrapolated
to ICU patients with other infection sites
or bloodstream infection or those who
are noncritically ill. In summary, our
findings were based on a diagnostic strat-
egy supported by quantitative tracheal as-
pirates or bronchoscopic respiratory sam-
ples obtained before the initiation of
antibiotic therapy in a cohort of patients
with two thirds of late-onset VAP epi-
sodes. De-escalation was performed in
38% of episodes with sensitive strains but
was not possible if the pathogen re-
mained unknown. The impact of quanti-
tative tracheal aspirate or bronchoscopic
techniques was comparable in terms of
mortality, and differences between bron-
choscopic and nonbronchoscopic rates of
de-escalation were minimal. Finally, our
findings suggest that the probability of
de-escalation is lower in the presence of
nonfermenting Gram-negative bacillus or
late-onset episodes.
Crit Care Med 2004 Vol. 32, No. 11
ACKNOWLEDGMENTS
We are indebted to Michael Maudsley
for editing the manuscript and to Montse
Olona for statistical advice.
REFERENCES
1. Campbell GD, Niederman MS, Broughton
WA, et al: Hospital-acquired pneumonia in
ventilated patients. Diagnosis, assessment of
severity, initial antimicrobial therapy and
preventive strategies. Am J Respir Crit Care
Med 1996; 153:1711–1725
2. Trouillet JL, Chastre J, Vuagnat A, et al:
Ventilator-associated pneumonia caused by
potentially drug-resistant bacteria. Am J Re-
spir Crit Care Med 1998; 157:531–539
3. Luna CM, Vujacich P, Niedermann MS, et al:
Impact of BAL data on the therapy and out-
come of ventilator-associated pneumonia.
Chest 1997; 111:676 – 685
4. Kollef MH, Rello J, Cammarata SK, et al:
Clinical cure and survival in Gram-positive
ventilator-associated pneumonia: Retrospec-
tive analysis of two double-blind studies
comparing linezolid with vancomycin. Inten-
sive Care Med 200; 30:388 –394
5. Wunderink RG, Rello J, Cammarata SK, et al:
Linezolid vs vancomycin: Analysis of two
double-blind studies of patients with methi-
cillin-resistant Staphylococcus aureus noso-
comial pneumonia. Chest 2003; 124:
1789 –1797
6. Fagon JY, Chastre J: Management of sus-
pected ventilator-associated pneumonia. Ann
Intern Med 2000; 19:133:1009
7. Paterson DL, Ko W-Ch, Von Gottberg A, et al:
Outcome of cephalosporin treatment for se-
rious infections due to apparently susceptible
organisms producing extended-spectrum
␤-lactamases: Implications for the clinical
microbiology laboratory. J Clin Microbiol
2001; 2206 –2212
8. Wunderink RG: Mortality and ventilator-
associated pneumonia. The best antibiotics
may be the least antibiotics. Chest 1993; 104:
993–995
9. Corbella X, Montero A, Pujol M, et al: Emer-
gence and rapid spread of carbapenem resis-
tance during a large and sustained hospital
outbreak of multiresistant Acinetobacter
baumannii. J Clin Microbiol 2000; 38:
4086 – 4095
10. Trouillet JL, Vuagnat A, Combes A, et al:
Pseudomonas aeruginosa ventilator-associ-
ated pneumonia: Comparison of episodes due
to piperacillin-resistant vs. piperacillin-
susceptible organisms. Clin Infect Dis 2002;
34:1047–1054
11. Karam GH, Niederman MS: How do we
achieve adequate therapy for severe infec-
tion? Crit Care Med 2003; 31:648 – 650
12. Höffken G, Niederman MS: Nosocomial
pneumonia: The importance of a de-escalat-
ing strategy for antibiotic treatment of pneu-
monia in the ICU. Chest 2002; 122:
2183–2196
13. Paterson DL, Rice LB: Empirical antibiotic
choice for the seriously ill patient: Are min-
imization of selection of resistant organisms
and maximization of individual outcome mu-
tually exclusive? Clin Infect Dis 2003; 36:
1006 –1012
14. Leone M, Bourgoin A, Cambon S, et al: Em-
pirical antimicrobial therapy of septic shock
patients: Adequacy and impact on the out-
come. Crit Care Med 2003; 31:462– 467
15. Bonten MJ, Bergmans DC, Stobberingh EE,
et al: Implementation of bronchoscopic tech-
niques in the diagnosis of ventilator-
associated pneumonia to reduce antibiotic
use. Am J Respir Crit Care Med 1997; 156:
1820 –1824
16. Ibrahim EH, Ward S, Sherman G, et al: Ex-
perience with a clinical guideline for the
treatment of ventilator-associated pneumo-
nia. Crit Care Med 2001; 29:1109 –1115
17. Rello J, Gallego M, Mariscal D, et al: The
value of routine microbial investigation in
ventilator-associated pneumonia. Am J Re-
spir Crit Care Med 1997; 156:196 –200
18. Rello J, Sa-Borges M, Correa H, et al: Varia-
tions in etiology of ventilator-associated
pneumonia across four treatment sites: Im-
plications for antimicrobial prescribing prac-
tices. Am J Respir Crit Care Med 1999; 160:
608 – 613
19. Ruiz M, Torres A, Ewig S, et al: Noninvasive
vs. invasive microbial investigation in venti-
lator-associated pneumonia: Evaluation of
outcome. Am J Respir Crit Care Med 2000;
162:119 –125
20. Sandiumenge A, Diaz E, Bodi M, et al: Ther-
apy of ventilator-associated pneumonia. A
patient-based approach based on the ten
rules of “The Tarragona Strategy.” Intensive
Care Med 2003; 29:876 – 883
21. Rello J, Diaz E: Pneumonia in the intensive
care unit. Crit Care Med 2003; 31:2544 –2551
22. Wunderink R, Mayhall G, Gibert C: Method-
ology for clinical investigation of ventilator-
associated pneumonia. Chest 1992; 102
(Suppl):580S–588S
23. Woodhead MA, Arrowsmith J, Chamberlain-
Webber R, et al: The value of routine micro-
bial investigation in community-acquired
pneumonia. Respir Med 1991; 85:313–317
24. Vallés J, Rello J, Ochagavia A, et al: Commu-
nity-acquired bloodstream infection in criti-
cally ill adult patients: Impact of shock and
inappropriate antibiotic therapy on survival.
Chest 2003; 123:1615–1624
25. Waterer JW, Jennings SG, Wunderink RG:
The impact of blood cultures on antibiotic
therapy in pneumococcal pneumonia. Chest
1999; 116:1278 –1281
26. Singh N, Rogers P, Atwood CW, et al: Short
course empiric antibiotic therapy for patients
with pulmonary infiltrates in the intensive
care unit. Am J Respir Crit Care Med 2000;
162:505–511
27. Jourdain B, Novara A, Joly-Guillou ML, et al:
Role of quantitative cultures of endotracheal
aspirates in the diagnosis of nosocomial
2189
 pneumonia. Am J Respir Crit Care Med 1995;
152:241–246
28. Chastre J, Wolff M, Fagon JY, et al: Compar-
ison of 8 vs 15 days of antibiotic therapy for
ventilator-associated pneumonia in adults: a
randomized trial. JAMA 2003; 290:
2588 –2598
29. Rello J, Ausina V, Ricart M, et al: Impact of
previous antimicrobial therapy on the etiol-
ogy and outcome of ventilator-associated
pneumonia. Chest 1993; 104:1230 –1235
30. Kollef MH, Ward S: The influence of mini-
BAL cultures on patient outcomes. Implica-
tion for the antibiotic management of venti-
lator-associated pneumonia. Chest 1998;
113:412– 420
31. Ibrahim EH, Sherman G, Ward S, et al: The
influence of inadequate antimicrobial treat-
ment of bloodstream infections on patient
outcomes in the ICU setting. Chest 2000;
118:146 –155
32. Alvarez-Lerma F: Modification of empiric an-
tibiotic treatment in patients with pneumo-
nia acquired in the intensive care unit. ICU-
Acquired Pneumonia Study Group. Intensive
Care Med 1996; 22:387-3-94
33. Heyland DK, Cook DJ, Griffith L, et al: The
attributable morbidity and mortality of ven-
tilator-associated pneumonia in the critically
ill patient. The Canadian Critical Trials
Group. Am J Respir Crit Care Med 1999;
159:1249 –1256
34. Namias N, Samiian L, Nino D, et al: Inci-
dence and susceptibility of pathogenic bacte-
ria vary between intensive care units within a
single hospital: Implications for empiric an-
tibiotic strategies. J Trauma 2000; 49:
638 – 645
35. Ioanas M, Ferrer M, Cavalcanti M, et al:
Causes and predictors of nonresponse to
treatment of intensive care unit-acquired
pneumonia. Crit Care Med 2004; 32:938 –945
36. Paterson DL, Ko WC, Von Gottberg A, et al:
International prospective study of Klebsiella
pneumoniae bacteremia: Implications of ex-
tended-spectrum beta-lactamase production
in nosocomial infections. Ann Intern Med
2004; 140:26 –32

ACCM Guidelines on SCCM Website

The Guidelines and Practice Parameters developed by the American College of Critical Care
Medicine are now available online at http://www.sccm.org/professional_resources/
guidelines/index.asp. The printed version of the Guidelines, provided in a binder, is also
available through the SCCM Bookstore, located at http://www.sccm.org/pubs/sccmbookstore.
html. Please watch the Website to stay updated on the ACCM Guidelines and Practice
Parameters.

2190 Crit Care Med 2004 Vol. 32, No. 11