Invasive and Noninvasive Strategies for Management of Suspected

Ventilator-Associated Pneumonia
A Randomized Trial
Jean-Yves Fagon, MD; Jean Chastre, MD; Michel Wolff, MD; Claude Gervais, MD;
Sylvie Parer-Aubas, MD; François Stéphan, MD; Thomas Similowski, MD; Alain Mercat, MD;
Jean-Luc Diehl, MD; Jean-Pierre Sollet, MD; and Alain Tenaillon, MD, for the VAP Trial Group*

Background: Optimal management of patients who are

clinically suspected of having ventilator-associated pneu-
monia remains open to debate.
T he diagnosis and treatment of ventilator-associ-
ated pneumonia, a nosocomial infection that
develops in mechanically ventilated patients and
Objective: To evaluate the effect on clinical outcome and causes considerable morbidity and mortality, remain
antibiotic use of two strategies to diagnose ventilator- a challenge (1–3). A presumptive clinical diagnosis
associated pneumonia and select initial treatment for this of pneumonia is often made when a patient devel-
condition.
ops a new radiographic infiltrate associated with
Design: Multicenter, randomized, uncontrolled trial. fever, leukocytosis, and purulent tracheal secretions
Setting: 31 intensive care units in France. and when microorganisms are isolated by nonquan-
Patients: 413 patients suspected of having ventilator- titative analysis of endotracheal aspirates (4). This
associated pneumonia. “clinical” approach leads to overestimation of the
Intervention: The invasive management strategy was
incidence of ventilator-associated pneumonia be-
based on direct examination of bronchoscopic protected cause cases of tracheobronchial colonization and
specimen brush samples or bronchoalveolar lavage sam- noninfectious processes mimicking it are included
ples and their quantitative cultures. The noninvasive (5–7). The nonspecificity of a strategy based on
(“clinical”) management strategy was based on clinical clinical evaluation has potentially deleterious con-
criteria, isolation of microorganisms by nonquantitative sequences: Many patients may receive unneeded
analysis of endotracheal aspirates, and clinical practice antibiotics; this exposes them to unnecessary toxic-
guidelines. ity, increases hospital costs, and favors the emer-
Measurements: Death from any cause, quantification of gence of resistant microorganisms. In addition, an-
organ failure, and antibiotic use at 14 and 28 days. tibiotic overuse in such patients delays diagnosis of
Results: Compared with patients who received clinical the true cause of fever and pulmonary infiltrate.
management, patients who received invasive manage- Concern about the inaccuracy of clinical ap-
ment had reduced mortality at day 14 (16.2% and 25.8%; proaches to diagnosis of ventilator-associated pneu-
difference, ⫺9.6 percentage points [95% CI, ⫺17.4 to ⫺1.8 monia led numerous investigators to postulate that
percentage points]; P ⫽ 0.022), decreased mean Sepsis-
“invasive” diagnostic methods, including quantita-
related Organ Failure Assessment scores at day 3 (6.1 ⫾ 4.0
and 7.0 ⫾ 4.3; P ⫽ 0.033) and day 7 (4.9 ⫾ 4.0 and
tive cultures of specimens obtained by using bron-
5.8 ⫾ 4.4; P ⫽ 0.043), and decreased antibiotic use (mean choscopic bronchoalveolar lavage or protected spec-
number of antibiotic-free days, 5.0 ⫾ 5.1 and 2.2 ⫾ 3.5; imen brush, could improve identification of patients
P ⬍ 0.001). At 28 days, the invasive management group with true ventilator-associated pneumonia and se-
had significantly more antibiotic-free days (11.5 ⫾ 9.0 lection of appropriate antibiotics (8 –10). However,
compared with 7.5 ⫾ 7.6; P ⬍ 0.001), and only multivari- these procedures require rigorous adherence to
ate analysis showed a significant difference in mortality bronchoscopic and microbiological techniques and
(hazard ratio, 1.54 [CI, 1.10 to 2.16]; P ⫽ 0.01). are not universally available; moreover, in the ab-
Conclusions: Compared with a noninvasive manage- sence of a definite gold standard for the diagnosis
ment strategy, an invasive management strategy was sig- of ventilator-associated pneumonia, the value of
nificantly associated with fewer deaths at 14 days, earlier such tests is uncertain, and their use in everyday
attenuation of organ dysfunction, and less antibiotic use in
practice remains controversial (4, 11, 12).
patients suspected of having ventilator-associated pneu-
monia.
To test the hypothesis that an invasive manage-
ment strategy is superior to a clinical, noninvasive
one in terms of improving clinical outcomes and
Ann Intern Med. 2000;132:621-630.
minimizing antibiotic use, we initiated a multicenter,
For author affiliations, current addresses, and contributions, see randomized, uncontrolled trial to compare these
end of text.
strategies in patients suspected of having ventilator-
*For members of the VAP Trial Group, see the Appendix. associated pneumonia. The primary end points were
© 2000 American College of Physicians–American Society of Internal Medicine 621
death from any cause, antibiotic use for any reason, Randomization and Data Collection
and quantification of organ failure during the first Patients were randomly assigned to receive the
14 days of follow-up. clinical or the invasive management strategy. Com-
puter-generated random-number tables were used
to assign patients in blocks of 8, with stratification
Methods according to treatment center.
At admission to the intensive care unit, we re-
Patient Selection and Study Design
corded each patient’s age; sex; severity of underly-
After obtaining approval of the institutional re- ing medical condition, stratified as rapidly fatal, ul-
view boards at each participating institution and timately fatal, or not fatal according to the criteria
informed consent from patients or their proxies, we of McCabe and Jackson (14); SAPS II score (range,
enrolled patients at 31 intensive care units. Inclu- 0 to 174; higher scores indicate more severe illness)
sion criteria were age older than 18 years; at least (13); Sepsis-related Organ Failure Assessment
48 hours of mechanical ventilation; and clinical sus- (SOFA) score (range, 0 to 24, with scores for each
picion of ventilator-associated pneumonia, defined organ system [respiration, coagulation, liver, cardio-
by new and persistent infiltrate on chest radiography vascular, central nervous system, and kidney] rang-
associated with at least one of the following: puru- ing from 0 [normal] to 4 [most abnormal]) (15); the
lent tracheal secretions, body temperature of at Organ Dysfunction and Infection (ODIN) score
least 38.3 °C, and leukocytosis. Exclusion criteria (range, 0 to 7, according to the presence or absence
were pregnancy; enrollment in another interven- of cardiovascular, respiratory, renal, hepatic, hema-
tional study; little chance of survival, defined by a tologic, and neurologic dysfunctions or infection)
Simplified Acute Physiologic Score II (SAPS II) of (16); classification as medical patient, surgical pa-
more than 65 points (corresponding to a probability tient with trauma, or surgical patient without
of death exceeding 77%) (13); and introduction or trauma according to the admitting diagnosis; and
modification of antibiotic therapy, instigated by new reason for initiating mechanical ventilation (17).
clinical symptoms, during the 3 days before collec- The following baseline variables were recorded
tion of respiratory samples. before randomization: SAPS II score; SOFA score;
For patients in the clinical management group, ODIN score; body temperature; leukocyte count;
the decision on whether to treat was based on clin- radiologic score (range, 0 to 12 according to the
ical evaluation and results of immediate microscopic density of radiologic infiltration) (18); ratio of the
examination of Gram-stained endotracheal aspi- partial pressure of arterial oxygen to the fraction of
rates. The results of the Gram stain and recommen- inspired oxygen (PaO2/FIO2); presence of shock, de-
dations of the American Thoracic Society on hospital- fined as systolic arterial pressure less than 90 mm
acquired pneumonia were used to guide the initial Hg with signs of peripheral hypoperfusion or need
choice of antibiotics (3). Results of qualitative aspi- for continuous infusion of vasopressor or inotropic
rate cultures were used to adjust the initial anti- agents (19); and presence of the acute respiratory
biotic regimen; when cultures were negative, no distress syndrome, defined as the presence of a gen-
treatment was given (Figure 1A). eralized pulmonary infiltrate and a lung injury score
The invasive strategy used fiberoptic bronchos- more than 2.5 (20); duration of previous mechanical
copy to obtain protected specimen brush samples or ventilation; and use or no use of antibiotics. These
bronchoalveolar lavage samples for direct micro- baseline variables (except SAPS II score) were mea-
scopic examination. Results of these examinations sured again 3, 7, 14, 21, and 28 days after the day of
were used to diagnose ventilator-associated pneu- inclusion. All infections requiring specific therapeu-
monia, to decide to treat, and to guide the initial tic measures during the first 3 days after inclusion
choice of antibiotics when specimens were positive. were recorded. Antibiotic use was recorded daily
Results of quantitative cultures were used to adjust until day 28.
therapy; treatment was discontinued when results
were negative, and use of antibiotics with narrower
spectra of activity was based on identification of and Specimen Collection and Microbiological
susceptibility-test results for pathogens cultured at Processing
significant concentrations (protected specimen brush Patients who received invasive management un-
sample that yielded ⱖ103 colony-forming units derwent fiberoptic bronchoscopy according to each
[CFU]/mL or bronchoalveolar lavage fluid sample center’s protocol. Premedication, use of a short-
that yielded ⱖ104 CFU/mL [4, 8 –10]) (Figure 1B). acting neuromuscular blocking agent, and adjust-
For both groups, the recommended duration of ment of FIO2 to 95% or more were recommended;
therapy for ventilator-associated pneumonia was 14 protected specimen brush, bronchoalveolar lavage,
days. or both were performed at the investigator’s discre-
622 18 April 2000 • Annals of Internal Medicine • Volume 132 • Number 8
Figure 1. Diagnostic and therapeutic strategy applied to patients managed with the clinical strategy (A) or invasive strategy (B). ATS ⫽
American Thoracic Society; BAL ⫽ bronchoalveolar lavage; PSB ⫽ protected specimen brush.

tion. Processing of microbiological specimens has into two samples: one for direct microscopic exam-
been described in detail elsewhere (10, 21). Briefly, ination of cytocentrifuge preparations after Gram or
recovered bronchoalveolar lavage fluid was divided modified Wright–Giemsa staining to determine the
18 April 2000 • Annals of Internal Medicine • Volume 132 • Number 8 623
Table 1. Admission and Baseline Characteristics of the Study Patients*

Characteristic Patients Who Received Invasive Patients Who Received Clinical

Management (n ⫽ 204) Management (n ⫽ 209)

Admission
Age, y 63 ⫾ 16 63 ⫾ 15
Sex, n (%)
Male 141 (69.1) 148 (70.8)
Female 63 (30.9) 61 (29.2)
McCabe–Jackson classification, n (%)
Nonfatal underlying disease 133 (65.2) 141 (67.5)
Ultimately fatal underlying disease 61 (29.9) 63 (30.1)
Rapidly fatal underlying disease 10 (4.9) 5 (2.4)
SAPS II score†‡ 44 ⫾ 15 42 ⫾ 14
SOFA score†‡ 7.8 ⫾ 4.1 7.1 ⫾ 3.9
ODIN score†‡ 2.1 ⫾ 1.1 1.9 ⫾ 1.0§
Classification of patients, n (%)
Medical 142 (69.6) 139 (66.5)
Surgical, no trauma 48 (23.5) 59 (28.2)
Surgical, trauma 14 (6.9) 11 (5.3)
Reason for mechanical ventilation, n (%)
Acute exacerbation of COPD 26 (12.7) 27 (12.9)
Acute respiratory failure 69 (33.8) 63 (30.1)
Postoperative respiratory failure 61 (29.9) 58 (27.8)
Drug overdose 3 (1.5) 1 (0.5)
Neurologic 37 (18.1) 46 (22.0)
Miscellaneous 8 (3.9) 14 (6.7)
Baseline
Duration of mechanical ventilation before study entry, d 10.4 ⫾ 10.2 10.7 ⫾ 10.0
Previous antimicrobial therapy, n (%) 105 (51.5) 103 (49.3)
SAPS II score†‡ 41 ⫾ 12 41 ⫾ 12
SOFA score†‡ 7.3 ⫾ 3.7 7.0 ⫾ 4.0
ODIN score†‡ 2.0 ⫾ 1.0 1.9 ⫾ 0.9
Site of organ failure, n (%)†
Cardiovascular system 84 (41.2) 83 (39.7)
Renal system 38 (18.6) 46 (22.0)
Central nervous system 65 (31.9) 40 (19.1)
Hepatic system 18 (8.8) 16 (7.7)
Hematologic system 11 (5.4) 5 (2.4)
Body temperature, °C 38.7 ⫾ 0.9 38.7 ⫾ 0.9
Leukocyte count, cells/mm3 15 190 ⫾ 7150 15 670 ⫾ 6800
Radiologic score† 5.2 ⫾ 2.7 5.0 ⫾ 2.6
PaO2/FIO2, mm Hg 221 ⫾ 86 215 ⫾ 93
Shock, n (%)† 74 (36.3) 81 (38.8)
Acute respiratory distress syndrome, n (%)† 26 (12.7) 22 (10.5)

* Data presented with a plus/minus sign are the mean ⫾ SD. COPD ⫽ chronic obstructive pulmonary disease; ODIN ⫽ Organ Dysfunction and Infection; PaO2/FIO2 ⫽ ratio of partial
pressure of arterial oxygen to the fraction of inspired oxygen; SAPS ⫽ Simplified Acute Physiology Score; SOFA ⫽ Sepsis-related Organ Failure Assessment.
† Described more fully in the Methods section.
‡ Higher values indicate greater severity.
§ P ⫽ 0.031 compared with the invasive therapy group. The groups did not significantly differ for any other characteristic.

percentages of cells containing intracellular bacteria, vortexed for 1 minute; Gram staining and qualita-
and the other for quantitative cultures. The tip of tive aerobic cultures were performed for all patients.
the protected specimen brush was cut, dropped into
1 mL of sterile water, and vortexed for 1 minute;
samples were examined directly and serially diluted Definitions
for culture. The number of bacteria in the original Inappropriate treatment, evaluated initially and
specimens was estimated by colony counts and is at 3 days, was defined as the use of antibiotics to
expressed as CFU/mL. Patients in the invasive treat- which at least one cultured isolate was resistant in
ment group were considered to have ventilator- vitro. For patients in the clinical management
associated pneumonia if more than 5% of the cells group, all pathogens grown in qualitative endotra-
in cytocentrifuge preparations of bronchoalveolar cheal aspirate cultures were considered for this
lavage fluid contained intracellular bacteria or at analysis; for patients in the invasive management
least one bacterial species grew at a significant con- group, only pathogens cultured at significant con-
centration from the protected specimen brush sam- centrations were taken into account.
ple (ⱖ103 CFU/mL) or from bronchoalveolar lavage Resistant bacteria were defined as ticarcillin-resis-
fluid (ⱖ104 CFU/mL) (10, 22). tant Pseudomonas aeruginosa, Acinetobacter bau-
In patients who received clinical treatment, en- mannii, and Stenotrophomonas maltophilia; extended-
dotracheal aspirates were collected sterilely by using spectrum ␤-lactamase–producing Enterobacteriaceae;
a suction catheter in a mucus collector; secretions and methicillin-resistant Staphylococcus aureus.
were aspirated without instilling saline. Aspirates were We calculated the number of antibiotic-free days
624 18 April 2000 • Annals of Internal Medicine • Volume 132 • Number 8
(days without antibiotic therapy) at 14 and 28 days Table 2. Features and Organisms Associated with
Ventilator-Associated Pneumonia*
after inclusion. For example, a patient who survived
28 days and received no antibiotics was assigned a Feature or Organism Patients Who Patients Who
value of 28. If antibiotics had been given for 10 days Received Invasive Received Clinical
Management Management
and the patient died on day 14, a value of 4 was (n ⫽ 204) (n ⫽ 209)
assigned. Using the same method, we calculated the
number of mechanical ventilation–free days. We Negative culture, n (%) 114 (55.9) 30 (14.4)
Monomicrobial pneumonia, n (%) 65 (31.9) 84 (40.2)
also calculated the number of antibiotics per day by Polymicrobial pneumonia, n (%) 25 (12.3) 95 (45.5)
dividing the sum of number of days of administra- Total number of pathogens, n 121 312
Bacilli, n (%)
tion of each antibiotic by the duration of survival at Pseudomonas aeruginosa 27 (22.3) 57 (18.3)
14 and 28 days. Haemophilus influenzae 9 (7.4) 12 (3.8)
Escherichia coli 6 (5.0) 23 (7.4)
Acinetobacter baumannii 6 (5.0) 11 (3.5)
Outcome Measures Enterobacter species 4 (3.3) 12 (3.8)
Proteus species 3 (2.5) 14 (4.5)
During their stay in the intensive care unit, many Serratia marcescens 3 (2.5) 7 (2.2)
patients experience an adverse event that can influ- Klebsiella species 2 (1.7) 11 (3.5)
Citrobacter species 1 (0.8) 7 (2.2)
ence their outcome. Because these complications Morganella morganii 1 (0.8) 3 (1.0)
are potential confounding factors in a study whose Moraxella species 1 (0.8) 1 (0.3)
Stenotrophomonas maltophilia 0 4 (1.3)
principal judgment criterion is outcome, we decided Corynebacterium 0 4 (1.3)
to evaluate mortality during the first 14 days be- Alcaligenes xylosoxidans 0 1 (0.3)
Cocci, n (%)
cause this period corresponds to that during which Staphylococcus aureus 20 (16.5) 40 (12.8)
ventilator-associated pneumonia has its maximal im- Streptococcus species 19 (15.7) 28 (9.0)
Neisseria species 7 (5.8) 6 (1.9)
pact on patient survival (2, 4, 21). In addition, nu- Streptococcus pneumoniae 3 (2.5) 10 (3.2)
merous factors come into play during the subse- Coagulase-negative staphylococci 3 (2.5) 17 (5.4)
Enterococcus species 1 (0.8) 6 (1.9)
quent 14 to 28 days. Thus, the primary end points Fungi, n (%) 5 (4.1) 38 (12.2)
of the study were mortality at 14 days; antibiotic-
free days at 14 days; and quantification of organ * Organisms shown are those that were isolated at significant concentrations from quan-
titative cultures of protected specimen brush samples (ⱖ103 colony-forming units/mL)
failure at 3, 7, and 14 days according to the SOFA or bronchoalveolar lavage samples (ⱖ104 colony-forming units/mL) in the invasive
management group and from qualitative cultures of endotracheal aspirates from the
and ODIN scores. Secondary end points were mor- clinical management group. Because of rounding, percentages do not always add up
tality at 28 days, antibiotic-free days at 28 days, to 100.

quantification of organ failure at 28 days (assessed

by using the SOFA and ODIN scores), mechanical Institute, Inc., Cary, North Carolina) was used for
ventilation–free days at 28 days, duration of stay in statistical analyses. A target sample size of 400 pa-
the intensive care unit, duration of hospital stay, tients was chosen to ensure, with a probability of
emergence of resistant bacteria, and emergence of 80%, the detection of a 10% difference between the
Candida species during 28 days. mortality rates of the two groups; a mortality rate of
30% was assumed for the clinical treatment group.
Statistical Analysis
Role of the Funding Sources
Data are presented as the mean (⫾SD). Survival
between groups was compared by using Kaplan– The funding agencies played no role in the de-
Meier curves and the log-rank test. Measurements sign of the study; collection, analysis, and interpre-
expressed as means or percentages were compared tation of data; or the decision to submit the paper
by using the t-test for continuous data and the chi- for publication.
square statistic with Yates correction for propor-
tions. All tests of statistical significance were two-
sided. We did not correct for multiple testing. An Results
intention-to-treat analysis was performed. To ana-
lyze the effect of management strategies on mortal- Comparison of Study Groups
ity at day 28, we used a Cox multivariate propor- From September 1996 to January 1998, 418 pa-
tional hazards model that included age, SAPS II tients were enrolled; 5 subsequently withdrew their
score, and McCabe–Jackson score at admission consent to receive a randomly assigned strategy and
(variables strongly associated with outcome of pa- for use of their data, leaving 413 patients (204 in
tients in the intensive care unit), and duration of the invasive management group and 209 in the clin-
mechanical ventilation before inclusion, ODIN ical management group). All variables correlated
score, PaO2/FIO2 ratio, and radiologic score mea- with mortality and morbidity, except ODIN score at
sured at inclusion (variables strongly associated with admission, were similar in the two groups (Table 1).
outcome of nosocomial pneumonia) as covariates Antibiotic management adhered to a prespecified
with the randomization group. SAS software (SAS protocol (Figure 1).
18 April 2000 • Annals of Internal Medicine • Volume 132 • Number 8 625
Table 3. Study Outcomes according to the Intention-to-Treat Analysis*

End Point Patients Who Received Patients Who Received Difference (95% CI) P Value
Invasive Management Clinical Management
(n ⫽ 204) (n ⫽ 209)

Primary
Mortality at 14 days, n (%) 33 (16.2) 54 (25.8) ⫺9.6 (⫺17.4 to ⫺1.8)† 0.022
Multiple organ dysfunction‡§
At 3 days
SOFA score 6.1 ⫾ 4.0 7.0 ⫾ 4.3 ⫺0.9 (⫺1.7 to ⫺0.1) 0.033
ODIN score 1.7 ⫾ 0.9 1.9 ⫾ 1.1 ⫺0.2 (⫺0.4 to ⫺0.05) 0.014
At 7 days
SOFA score 4.9 ⫾ 4.0 5.8 ⫾ 4.4 ⫺0.9 (⫺1.8 to ⫺0.03) 0.043
ODIN score 1.4 ⫾ 1.0 1.6 ⫾ 1.1 ⫺0.2 (⫺0.4 to 0.02) 0.082
At 14 days
SOFA score 3.9 ⫾ 4.1 4.3 ⫾ 4.3 ⫺0.4 (⫺1.3 to 0.6) ⬎0.2
ODIN score 1.2 ⫾ 1.2 1.2 ⫾ 1.2 ⫺0.03 (⫺0.3 to 0.2) ⬎0.2
Antibiotic-free days at 14 days, d‡ 5.0 ⫾ 5.1 2.2 ⫾ 3.5 2.8 (1.9 to 3.6) ⬍0.001
Antibiotics per day at 14 days, n 1.2 ⫾ 0.8 1.5 ⫾ 0.7 ⫺0.3 (⫺0.5 to ⫺0.2) ⬍0.001
Antibiotic-treatment days at 14 days, d 8.7 ⫾ 5.4 10.9 ⫾ 4.5 ⫺2.2 (⫺3.2 to ⫺1.2) ⬍0.001
Secondary
Mortality at 28 days, n (%) 63 (30.9) 81 (38.8) ⫺7.9 (⫺17.0 to 1.2) 0.099
Multiple organ dysfunction at 28 days‡§
SOFA score 3.1 ⫾ 3.4 3.1 ⫾ 3.8 ⫺0.02 (⫺1.2 to 1.1) ⬎0.2
ODIN score 1.0 ⫾ 1.0 1.0 ⫾ 1.0 ⫺0.06 (⫺0.4 to 0.3) ⬎0.2
Antibiotic-free days at 28 days, d‡ 11.5 ⫾ 9.0 7.5 ⫾ 7.6 ⫺3.9 (⫺5.5 to ⫺2.3) ⬍0.001
Antibiotics per day at 28 days, n 1.0 ⫾ 1.8 1.3 ⫾ 0.7 ⫺0.3 (⫺0.45 to ⫺0.16) ⬍0.001
Antibiotic-treatment days at 28 days, d 12.8 ⫾ 8.5 14.9 ⫾ 7.9 ⫺2.1 (⫺3.7 to ⫺0.5) 0.009
Duration of intensive care unit stay, d 19.3 ⫾ 9.0 17.6 ⫾ 9.4 1.5 (⫺0.3 to 3.2) 0.11
Duration of hospital stay, d 26.7 ⫾ 23.9 25.1 ⫾ 28.5 1.6 (⫺0.3 to 3.4) ⬎0.2
Mechanical ventilation–free days, d‡ 7.8 ⫾ 9.8 7.0 ⫾ 9.4 0.8 (⫺1.0 to 2.9) ⬎0.2
Emergence of resistant bacteria, n (%) 125 (61.3) 125 (59.8) 1.5 (⫺7.9 to 10.9) ⬎0.2
Emergence of Candida species, n (%) 23 (11.3) 47 (22.5) ⫺11.2 (⫺18.3 to ⫺4.1) 0.0025

* Unless otherwise indicated, data are presented as the mean ⫾ SD. ODIN ⫽ Organ Dysfunction and Infection; SOFA ⫽ Sepsis-related Organ Failure Assessment.
† Expressed as percentage points.
‡ Described more fully in the Methods section.
§ Higher values indicate greater severity.

In the invasive management group, protected rate cultures were positive in 179 of 209 patients
specimen brush (n ⫽ 67) or bronchoalveolar lavage (85.6%). All identified microorganisms are reported
(n ⫽ 31) was used, alone or together (n ⫽ 106). Mi- (Table 2).
crobial cultures were positive in 64 of 173 (37.0%)
protected specimen brush samples and 46 of 137
End Point Analyses
(33.6%) bronchoalveolar lavage samples, for a total
of 90 cases of bacteriologically confirmed ventilator- At 14 days, 33 of 204 patients in the invasive
associated pneumonia in 204 patients (44.1%). In management group and 54 of 209 patients in the
the clinical management group, endotracheal aspi- clinical management group had died (16.2% and
25.8%, respectively; difference, ⫺9.6 percentage
points [95% CI, ⫺17.4 to ⫺1.8 percentage points];
P ⫽ 0.022) (Table 3). As seen in Figure 2, survival
curves remained parallel once this difference be-
tween groups was established. The mean SOFA
score was significantly lower in the invasive manage-
ment group than the clinical management group at
3 and 7 days but not at 14 days. The invasive
management group had significantly more antibiot-
ic-free days and received significantly fewer antibi-
otics per day (Table 3).
At 28 days, no significant differences in survival,
number of organ failures, duration of ICU stay,
mechanical ventilation–free days, or emergence of
resistant bacteria were seen between groups (Table
3). However, multivariate proportional hazards re-
gression analysis of the prognostic role of age,
SAPS II score, and McCabe–Jackson classification
Figure 2. Actuarial 28-day survival among 413 patients assigned
to the invasive (solid line) or clinical (dashed line) management
at admission; duration of mechanical ventilation be-
strategy. P ⫽ 0.07 for difference between groups (log-rank test). fore inclusion; and ODIN score, PaO2/FIO2 ratio,
626 18 April 2000 • Annals of Internal Medicine • Volume 132 • Number 8
and radiologic score at baseline showed that mor-
tality at 28 days was significantly higher in the clin-
ical management group (hazard ratio, 1.54 [CI, 1.10
to 2.16]; P ⫽ 0.01) (Table 4).
The invasive management group had significantly
more antibiotic-free days and fewer antibiotics per
day at 28 days (Table 3). Twenty-nine patients
(14%) in this group received no antibiotics up to
day 28 compared with only 4 (2%) in the clinical
management group (P ⬍ 0.001). Figure 3 shows the
use of the 14 most commonly prescribed antibiotics
in patients of both groups. Colonization or infection
with Candida species was documented in 23 patients
in the invasive management group and 47 patients
in the clinical management group (11.3% and
22.6%, respectively; P ⫽ 0.0025). Figure 3. Comparison of clinical (black circles) and invasive (white
Significant differences between the primary and circles) management strategies for the number of days of survival
without use of 1 of the 14 most commonly prescribed antibiotics.
secondary end points in the entire study sample For every antibiotic except imipenem, patients in the invasive management
were also observed for the subgroup of 318 patients group had significantly more antibiotic-free days. Error bars represent 95%
CIs.
with suspected late-onset pneumonia, defined as a
period between intubation and inclusion that ex-
ceeded 4 days. In contrast, for patients with early- Antibiotic Treatment
onset pneumonia, only the number of antibiotic-free Of the 114 patients in the invasive management
days at 14 days was significantly higher in the inva- group who had negative quantitative cultures, 97 did
sive management group than in the clinical manage- not receive antibiotics immediately, compared with
ment group (3.63 and 1.86 days, respectively; P ⫽ 18 of 30 patients in the clinical management group
0.024). In the subgroup of 208 patients who had re- who had negative endotracheal aspirate cultures
ceived previous antibiotic therapy, the only significant (85.1% and 60.0%, respectively; P ⫽ 0.0017) (Figure
differences were lower rates of antibiotic use and 1). Respective mortality rates at 14 days for these
mortality in the invasive management group (hazard groups of initially untreated patients were similar
ratio, 1.7 [95% CI, 1.03 to 2.71]; P ⫽ 0.039). The (15 of 97 patients [15.5%] and 3 of 18 [16.7%]).
effects of the invasive strategy on outcome were not For patients whose quantitative bronchoscopic
modified after stratification by intensive care unit. specimen cultures (invasive management group) or
qualitative endotracheal aspirate cultures (clinical
management group) grew pathogens, analysis of ad-
equacy of prescribed antibiotics and susceptibility-
Table 4. Cox Regression Analysis Hazard Ratios for test results showed that 1 patient in the former
Death at 14 and 28 Days*
group and 24 patients in the latter group had at
Covariate Hazard Ratio (95% CI) least one pathogen that was resistant to the anti-
14 Days 28 Days
biotic initially prescribed (P ⬍ 0.001). Thirty-two
percent of these 25 patients died compared with
Age, per 10-year increase 1.17 (0.98 –1.40) 1.23 (1.06 –1.42) 20.4% of the 388 patients who received appropriate
McCabe–Jackson classification
Nonfatal underlying disease 0.44 (0.13–1.46) 0.56 (0.22–1.44)
initial therapy (P ⬎ 0.2).
Ultimately fatal underlying Of the 24 patients in the clinical management
disease
Rapidly fatal underlying disease
0.64 (0.22–1.83)
0.83 (0.29 –2.37)
0.73 (0.31–1.70)
0.94 (0.40 –2.23)
group who initially received inappropriate antibiot-
SAPS II score at admission, per ics, 22 (92%) had been treated according to the
10-point increase
Duration of mechanical ventilation
0.91 (0.77–1.07) 0.94 (0.83–1.07)
American Thoracic Society guidelines; 10 received
before entry, per 5-day imipenem and an aminoglycoside and 6 received an
increase
Radiologic score at baseline, per
0.97 (0.87–1.08) 1.01 (0.94 –1.10)
antipseudomonal ␤-lactam, aminoglycoside, and
1-point increase 1.03 (0.94 –1.12) 1.02 (0.95–1.09) vancomycin. These 24 instances of inappropriate use
PaO2/FIO2 ratio at baseline, per
50 –mm Hg decrease 1.06 (0.93–1.20) 1.04 (0.94 –1.15)
of antibiotics were attributed to methicillin-resistant
ODIN score at baseline, per S. aureus (n ⫽ 10), multiresistant P. aeruginosa (n ⫽
1-point increase
Clinical strategy
2.10 (1.69 –2.61)
1.96 (1.26 –3.05)
1.77 (1.48 –2.12)
1.54 (1.10 –2.16)
8), A. baumannii (n ⫽ 4), and class 1 cephalospori-
nase-producing Enterobacteriaceae (n ⫽ 2). At 3
* ODIN ⫽ Organ Dysfunction and Infection; PaO2/FIO2 ⫽ ratio of partial pressure of days, after culture results had been obtained, no
arterial oxygen to the fraction of inspired oxygen; SAPS ⫽ Simplified Acute Physiology
Score. patient in the invasive management group and 2
18 April 2000 • Annals of Internal Medicine • Volume 132 • Number 8 627
patients in the clinical management group were still men cultures and did not receive antibiotics imme-
receiving inappropriate therapy. At 14 days, no pa- diately, the mortality rate was low (15.5%) and was
tient in the invasive management group and 8 pa- similar to that in patients who received clinical man-
tients (33.3%) in the clinical management group agement and had negative qualitative endotracheal
who initially received inappropriate antibiotics had aspirate cultures (16.7%). Results of studies by Ster-
died (P ⬎ 0.2). ling and associates (26), who conducted a decision
Between days 1 and 3, 22 infections requiring analysis comparing clinical and invasive strategies,
specific therapeutic measures occurred in the inva- and Heyland and colleagues (27), who conducted a
sive management group; these included 10 opera- prospective, nonrandomized trial comparing clinical
tion-site infections, 7 catheter-related infections, 3 and invasive strategies, support our observations of
cases of nosocomial pneumonia, and 2 cases of max- improved management and less antibiotic use in the
illary sinusitis. In contrast, only 5 such infections invasive management group.
occurred in the clinical management group (2 oper- Our study was limited by uncertainty about the
ation-site infections, 2 catheter-related infections, potential effect of its unblinded design. Theoreti-
and 1 case of sinusitis) (P ⬍ 0.001). cally, comparison of two strategies precludes the use
of a protocol in which the investigators are unaware
Discussion of the strategy assignments. Nevertheless, our con-
certed efforts to standardize intensive care and use
We found that in patients suspected of having of rigorous criteria to evaluate patient outcome
ventilator-associated pneumonia, an invasive strat- probably minimized additional bias due to differ-
egy based on use of fiberoptic bronchoscopy to di- ences in the management of severely ill patients
rectly sample a suspected area and quantitative cul- (Figure 1). The second limitation was use of two
tures to distinguish infecting pathogens from different diagnostic techniques to obtain secretions
colonizing microorganisms improved the survival rate, for quantitative cultures from patients assigned to
decreased antibiotic use, and was associated with the invasive strategy (11). Protected specimen brush
fewer organ failures 3 and 7 days after inclusion. and bronchoalveolar lavage have been extensively
Only a few studies have assessed the impact of evaluated, and the ranges of specificities and sensi-
diagnostic strategy on antibiotic use and outcome in tivities found for these techniques have suggested
patients suspected of having nosocomial pneumonia similar overall accuracy (12). The rates of positivity
(23–27). On the basis of a randomized study evalu- observed in our study, which were similar for both
ating 51 patients, Sanchez-Nieto and colleagues (24) techniques, suggest that the choice of one technique
reported significantly more modifications of the ini- or the other probably did not significantly influence
tial antimicrobial therapy for patients managed with the results. Third, in the clinical group, qualitative
an invasive strategy than in those managed with a rather than quantitative cultures of endotracheal as-
clinical strategy, but they found no significant influ- pirates were chosen because therapy based on clin-
ence on mortality. Unfortunately, that study had ical evaluation, nonquantitative cultures of endotra-
several limitations related to the small sample size, cheal aspirates, and knowledge of epidemiologic
the unbalanced distribution of factors pertinent to data remains the working strategy of the over-
ventilator-associated pneumonia mortality despite whelming majority of physicians (4).
randomization (higher frequencies of P. aeruginosa Finally, crude mortality rates differed significantly
infections and inappropriate initial treatments in the between the two groups at 14 days, when the death
invasive group), and lack of a coherent management rate in the clinical management group significantly
protocol in the invasive management group (initia- exceeded that in the invasive management group by
tion of treatment based on clinical evaluation and about 9%; this difference persisted over the follow-
continuation of antibiotic therapy in all patients de- ing 2 weeks but was no longer significant at 28 days.
spite negative cultures) (24). Concerning use of an- Nonetheless, the strategy applied continued to affect
tibiotics, in a study of 138 patients evaluated by the outcome: Eighteen more deaths occurred in the
collection of bronchoscopic specimens, Bonten and clinical management group, and the Cox propor-
coworkers (25) showed that antibiotic therapy can tional hazards model identified a higher mortality
be stopped in patients with negative quantitative rate in this group.
cultures with no adverse effect on recurrence of The lower mortality rate seen in the invasive
ventilator-associated pneumonia or mortality (25). management group might be explained by at least
Other researchers have also concluded that antibi- three different factors. First, the antibiotics initially
otic therapy can be safely stopped in patients with prescribed were more often appropriate in this
negative quantitative cultures (7, 27). In our study, group; there were significantly fewer inappropriate
among patients in the invasive management group treatments, a factor known to be associated with
who had negative quantitative bronchoscopic speci- lower mortality (1, 2, 23, 28, 29). However, inappro-
628 18 April 2000 • Annals of Internal Medicine • Volume 132 • Number 8
priate use of antibiotics was particularly low in our Appendix
clinical management group (13%) compared with
The following persons and institutions participated in
previous studies (30, 31), probably because investi-
the VAP Trial.
gators were required to carefully follow the recom-
Scientific Committee: M.O. Carrère, J. Chastre (Chair),
mendations of the American Thoracic Society. Only G. Duru, C. Gervais, C. Gibert, L. Gutmann, M. Safar.
one patient in the invasive management group with Trial Coordinating Center: Hôpital Broussais, Paris,
a positive bronchoscopic specimen was initially France. J.-Y. Fagon, C. Brun, I. Gautier.
treated incorrectly, probably because of the addi- Statistical and Data Management Center: Centre Na-
tional information obtained by examination of spec- tional de la Recherche Scientifique, Unité de Recherche
imens obtained directly (32). Associee 936, Lyon, France.
Second, patients who underwent bronchoscopy Participating investigators and centers: E. Azoulay and
ended up receiving fewer antibiotics, and antibiotic B. Schlemmer, Hôpital Saint-Louis, Paris; F. Blot, Institut
Gustave Roussy, Villejuif; C. Brun-Buisson and Y. Lefort,
therapy was discontinued in more of them. This
Hôpital Henri Mondor, Créteil; F. Bruneel and M. Wolff,
lowering of the risk for inadequate or unnecessarily
Hôpital Bichat–Claude Bernard, Paris; J. Chastre, J.L.
administered empirical therapy has major advan- Trouillet, L. Fierobe, and J.L. Dumoulin, Hôpital Bichat,
tages, namely avoidance of potentially harmful side Paris; J.-Y. Fagon and A. Novara, Hôpital Broussais,
effects and reduction of selection pressure, thereby Paris; C. Crombe and Y. Sucin, Centre Hospitalier Arles,
limiting the emergence of resistant microorganisms Arles; C. Delafosse and T. Similowski, Groupe Hospi-
and the corresponding heightened risk for superin- talier Pitié-Salpêtrière, Paris; J.L. Diehl and J. Labrousse,
fection known to adversely affect patient outcomes Hôpital Boucicaut, Paris; A. de Lassence and E. Le
(33, 34). Almost all reports emphasize that better Mière, Hôpital Louis Mourier, Colombes; Y. Domart and
antibiotic control programs to limit bacterial resis- B. Darchy, Centre Hospitalier de Compiègne, Comp-
iègne; F. Fraisse, Hôpital Delafontaine, Saint-Denis; B.
tance are urgently needed in this setting and that
Garrigues and J.L. Legrand, Centre Hospitalier du Pays
patients without true infections should not receive
d’Aix, Aix-en-Provence; M. Garrouste-Orgeas and B. Mis-
antibiotics. set, Hôpital Saint-Joseph, Paris; C. Gervais, R. Cohendy,
Finally, probably the most important risk of not C. Bengler, and P. Poupard, Hôpital Gaston Doumergue,
performing bronchoscopy is that another site of si- Nı̂mes; M. Hira and R. Robert, Centre Hospitalier Uni-
multaneous or subsequent infection may be missed. versitaire Poitiers, Poitiers; L. Holzapfel and G. Demingeon,
The major benefit of a negative bronchoscopy spec- Hôpital de Fleyriat, Bourg-en-Bresse; J. Lambert and T.
imen may be to direct attention away from the lungs Finge, Centre Hospitalier Alès, Alès; C. Lamer and Y.
as the source of fever, as demonstrated by the 22 Tric, Institut Mutualiste Montsouris (Centre Choisy), Par-
infections at other sites documented between days 1 is; A. Mercat and C. Richard, Hôpital Bicêtre, Le Krem-
lin-Bicêtre; G. Meyer and M. Djibre, Hôpital Laënnec,
and 3 in patients in the invasive management group
Paris; S. Parer-Aubas, Hôpital Lapeyronie, Montpellier;
(compared with 5 patients in the clinical manage-
J.F. Perrigault, Hôpital Saint Eloi, Montpellier; M. Slama
ment group). Many hospitalized patients with neg- and C. Galy, Hôpital Nord, Amiens; J.P. Sollet and H.
ative bronchoscopic specimen cultures have other Mentec, Hôpital Victor Dupouy, Argenteuil; F. Stéphan,
potential sites of infection that can more readily be Hôpital Tenon, Paris; A. Tenaillon and D. Perrin-Gal-
identified in the absence of antibiotic interference chadoat, Hôpital Louise Michel, Evry; G. Trouillet, Hô-
(2, 12, 35, 36). Delaying diagnosis or definitive pital René Dubos, Pontoise; D. Villers and E. Bironneau,
treatment of the true site of infection may lead to Hôtel Dieu, Nantes; M. Wysocki and M.A. Wolff, Institut
prolonged antibiotic therapy, more antibiotic-associ- Mutualiste Montsouris (Centre Jourdan), Paris.
ated complications, and induction of additional or- From Hôpital Broussais, Hôpital Bichat–Claude Bernard, Hôpi-
gan dysfunction (2, 7, 37). tal Tenon, Hôpital Pitié-Salpêtrière, and Hôpital Boucicaut, Par-
From this randomized study of two strategies to is; Centre Hospitalier Universitaire Nı̂mes, Nı̂mes; Centre Hos-
pitalier Universitaire Montpellier, Montpellier; Hôpital Bicêtre,
manage patients in whom ventilator-associated pneu- Le Kremlin-Bicêtre; Centre Hospitalier Victor Dupouy, Argen-
monia was suspected, we conclude that a strategy teuil; and Centre Hospitalier Louise Michel, Evry, France.
based on quantitative bronchoscopic specimen cul-
Grant Support: In part by the Société de Réanimation de Langue
tures has beneficial effects: improved early survival, Française and the Délégation à la Recherche Clinique, Assis-
fewer early organ failures, and less antibiotic use. It tance Publique–Hôpitaux de Paris.
provides arguments to stop giving antibiotics to pa-
Requests for Single Reprints: Jean-Yves Fagon, MD, Service de
tients without adequately identified microorganisms. Réanimation Médicale, Hôpital Broussais, 96 rue Didot, 75674
Because the invasive treatment strategy prescribes Paris Cedex 14, France.
fewer antibiotics, it limits their overuse and provides
Requests To Purchase Bulk Reprints (minimum, 100 copies): Bar-
clearer guidelines for the management of ventilated bara Hudson, Reprints Coordinator; phone, 215-351-2657; e-mail,
patients suspected of having nosocomial pneumonia. [email protected].

18 April 2000 • Annals of Internal Medicine • Volume 132 • Number 8 629

Current Author Addresses: Dr. Fagon: Service de Réanimation Am J Respir Crit Care Med. 1995;151:1878-88.
Médicale, Hopital Europeen Georges Pompidou, 20 rue Leblanc, 10. Chastre J, Fagon JY, Bornet-Lesco M, Calvat S, Dombret MC, al Khani
R, et al. Evaluation of bronchoscopic techniques for the diagnosis of noso-
75015 Paris, France.
comial pneumonia. Am J Respir Crit Care Med. 1995;152:231-40.
Dr. Chastre: Service de Réanimation Médicale, Hôpital Bichat– 11. Niederman MS, Torres A, Summer W. Invasive diagnostic testing is not
Claude Bernard, 46 rue Henri Huchard, 75018 Paris, France. needed routinely to manage ventilator-associated pneumonia. Am J Respir
Dr. Wolff: Clinique de Réanimation des Maladies Infectieuses, Crit Care Med. 1994;150:565-9.
Hôpital Bichat–Claude Bernard, 46 rue Henri Huchard, 75018 12. Chastre J, Fagon JY. Invasive diagnostic testing should be routinely used to
Paris, France. manage ventilated patients with suspected pneumonia. Am J Respir Crit Care
Med. 1994;150:570-4.
Dr. Gervais: Réanimation Médicale, CHU Nı̂mes, 5 rue Hoche, 13. Le Gall JR, Lemeshow S, Saulnier F. A new Simplified Acute Physiology
30029 Nı̂mes Cedex, France. Score (SAPS II) based on a European/North American multicenter study.
Dr. Parer-Aubas: DAR A–Réanimation, CHU Montpellier, Hô- JAMA. 1993;270:2957-63.
pital Lapeyronie, 371 avenue du Doyen Gaston Girard, 34295 14. McCabe WR, Jackson GG. Gram-negative bacteremia, I. Arch Intern Med.
Montpellier Cedex 5, France. 1982;110:847-64.
15. Vincent JL, Moreno R, Takala J, Willatts S, De Mendonca A, Bruining
Dr. Stéphan: Service d’Anesthesie–Réanimation, Hôpital Henri H, et al. The SOFA (Sepsis-related Organ Failure Assessment) score to de-
Mondor, 51 avenue du Maréchal de Lattre de Tassigny, 94010 scribe organ dysfunction/failure. On behalf of the Working Group on Sepsis-
Créteil Cedex, France. Related Problems of the European Society of Intensive Care Medicine. Inten-
Dr. Similowski: Service de Réanimation Pneumologique, Hôpital sive Care Med. 1996;22:707-10.
Pitié-Salpêtrière, 47-83 Boulevard de l’Hôpital, 75651 Paris Ce- 16. Fagon JY, Chastre J, Novara A, Medioni P, Gibert C. Characterization of
intensive care unit patients using a model based on the presence or absence
dex 13, France. of organ dysfunctions and/or infection: the ODIN model. Intensive Care Med.
Dr. Mercat: Service de Réanimation Médicale, Hôpital Bicêtre, 78 1993;19:137-44.
rue du Général Leclerc, 94275 Le Kremlin-Bicêtre Cedex, France. 17. Zwillich CW, Pierson DJ, Creagh CE, Sutton FD, Shatz E, Petty TL.
Dr. Diehl: Service de Réanimation Médicale, Hopital Europeen Complications of assisted ventilation. A prospective study of 354 consecutive
Georges Pompidou, 20 rue Leblanc, 75015 Paris, France. episodes. Am J Med. 1974;57:161-70.
18. Weinberg PR, Matthay MA, Webster RO, Roskos KV, Golstein IM,
Dr. Sollet: Service de Réanimation Médicale, Hôpital Murray JF. Biologically active products of complement and acute lung injury
d’Argenteuil, 69 rue du Lieutenant Colonel Prudhon, 95100 Ar- in patients with the sepsis syndrome. Am Rev Respir Dis. 1984;130:791-6.
genteuil, France. 19. Bone RC, Balk RA, Cerra FB, Dellinger RP, Fein AM, Knaus WA, et al.
Dr. Tenaillon: Service de Réanimation Médicale, Centre Hospi- Definitions for sepsis and organ failure and guidelines for the use of innova-
talier Louise Michel, Quartier du Canal, 91014 Evry Cour- tive therapies in sepsis. The ACCP/SCCm Consensus Conference Committee.
American College of Chest Physicians/Society of Critical Care Medicine.
couronnes Cedex, France.
Chest. 1992;101:1644-55.
20. Murray JF, Matthay MA, Luce JM, Flick MR. An expanded definition of
Author Contributions: Conception and design: J.-Y. Fagon, the adult respiratory distress syndrome. Am Rev Respir Dis. 1988;138:720-3.
J. Chastre, C. Gervais. 21. Chastre J, Fagon JY, Trouillet JL. Diagnosis and treatment of nosocomial
Analysis and interpretation of the data: J.-Y. Fagon, J. Chastre. pneumonia in patients in intensive care units. Clin Infect Dis. 1995;21(Suppl
Drafting of the article: J.-Y. Fagon, J. Chastre. 3):S226-S237.
22. Meduri GU, Chastre J. The standardization of bronchoscopic techniques for
Critical revision of the article for important intellectual con- ventilator-associated pneumonia. Chest. 1992;102(Suppl 1):557S-564S.
tent: J.-Y. Fagon, J. Chastre, M. Wolff, C. Gervais, S. Parer- 23. Luna CM, Vujavich P, Niederman MS, Vay C, Gherardi C, Matera J, et
Aubas, F. Stéphan, T. Similowski, A. Mercat, J.-L. Diehl, J.-P. al. Impact of BAL data on the therapy and outcome of ventilator-associated
Sollet, A. Tenaillon. pneumonia. Chest. 1997;111:676-85.
Final approval of the article: J.-Y. Fagon, J. Chastre, M. Wolff, 24. Sanchez-Nieto JM, Torres A, Garcia-Cordoba F, El-Ebiary M, Carrillo A,
Ruiz J, et al. Impact of invasive and noninvasive quantitative culture sam-
C. Gervais, S. Parer-Aubas, F. Stéphan, T. Similowski, A. Mer- pling on outcome of ventilator-associated pneumonia: a pilot study. Am J
cat, J.-L. Diehl, J.-P. Sollet, A. Tenaillon. Respir Crit Care Med. 1998;157:371-6.
Provision of study materials or patients: J.-Y. Fagon, J. Chas- 25. Bonten MJ, Bergmans DC, Stobberingh EE, van der Geest S, De Leeuw
tre, M. Wolff, C. Gervais, S. Parer-Aubas, F. Stéphan, T. PW, van Tiel FH, et al. Implementation of bronchoscopic techniques in the
Similowski, A. Mercat, J.-L. Diehl, J.-P. Sollet, A. Tenaillon. diagnosis of ventilator-associated pneumonia to reduce antibiotic use. Am J
Respir Crit Care Med. 1997;156:1820-4.
Statistical expertise: J.-Y. Fagon, J. Chastre. 26. Sterling TR, Ho EJ, Brehm WT, Kirkpatrick MB. Diagnosis and treatment
Obtaining of funding: J.-Y. Fagon, J. Chastre. of ventilator-associated pneumonia—impact on survival. A decision analysis.
Administrative, technical, or logistic support: J.-Y. Fagon, J. Chest. 1996;110:1025-34.
Chastre. 27. Heyland DK, Cook DJ, Marshall J, Heule M, Guslits B, Lang J, et al. The
Collection and assembly of data: J.-Y. Fagon, J. Chastre. clinical utility of invasive diagnostic techniques in the setting of ventilator-
associated pneumonia. Canadian Critical Care Trials Group. Chest. 1999;115:
1076-84
28. Kollef MH, Silver P, Murphy DM, Trovillion E. The effect of late-onset
References ventilator-associated pneumonia in determining patient mortality. Chest.
1995;108:1655-62.
1. Craven DE, Steger KA, Barber TW. Preventing nosocomial pneumonia: 29. Rello J, Ausina V, Ricart M, Castella J, Prats G. Impact of previous
state of the art and perspective for the 1990s. Am J Med. 1991;91:44S-53S. antimicrobial therapy on the etiology and outcome of ventilator-associated
2. Wunderink RG. Mortality and the diagnosis of ventilator-associated pneu- pneumonia. Chest. 1993;104:1230-5.
monia: a new direction. Am J Respir Crit Care Med. 1998;157:349-50. 30. Alvarez-Lerma F. Modification of empiric antibiotic treatment in patients
3. Baker AM, Meredith JW, Haponik EF. Pneumonia in intubated trauma with pneumonia acquired in the intensive care unit. ICU-Acquired Pneumonia
patients. Microbiology and outcomes. Am J Respir Crit Care Med. 1996;153: Study Group. Intensive Care Med. 1996;22:387-94.
343-9. 31. Kollef MH, Ward S. The influence of mini-BAL cultures on patients’ out-
4. Hospital-acquired pneumonia in adults: diagnosis, assessment of severity, ini- comes: implications for the antibiotic management of ventilator-associated
tial antimicrobial therapy, and preventing strategies. A consensus statement, pneumonia. Chest. 1998;113:412-20.
American Thoracic Society, November 1995. Am J Respir Crit Care Med. 32. Chastre J, Fagon JY, Soler P, Bornet M, Domart Y, Trouillet JL, et al.
1996;153:1711-25. Diagnosis of nosocomial bacterial pneumonia in intubated patients undergo-
5. Fagon JY, Chastre J, Hance AJ, Domart Y, Trouillet JL, Gibert C. ing ventilation: comparison of the usefulness of bronchoalveolar lavage and
Evaluation of clinical judgment in the identification and treatment of nosoco- the protected specimen brush. Am J Med. 1988;85:499-506.
mial pneumonia in ventilated patients. Chest. 1993;103:547-53. 33. Meduri GU, Mauldin GL, Wunderink RG, Leeper KV Jr, Jones CB,
6. Helling TS, Van Way C 3d, Krantz S, Bertram K, Stewart A. The value Tolley E, et al. Causes of fever and pulmonary densities in patients with clinical
of clinical judgment in the diagnosis of nosocomial pneumonia. Am J Surg. manifestations of ventilator-associated pneumonia. Chest. 1994;106:221-35.
1996;171:570-5. 34. Chastre J, Trouillet JL, Vuagnat A, Joly-Guillou ML, Clavier H, Dombret
7. Croce MA, Fabian TC, Shaw B, Stewart RM, Pritchard FE, Minard G, et MC, et al. Nosocomial pneumonia in patients with acute respiratory distress
al. Analysis of charges associated with diagnosis of nosocomial pneumonia: syndrome. Am J Respir Crit Care Med. 1998;157:1165-72.
can routine bronchoscopy be justified? J Trauma. 1994;37:721-7. 35. Meyer KS, Urban C, Eagan JA, Berger BJ, Rahal JJ. Nosocomial outbreak
8. Torres A, el-Ebiary M, Padro J, Gonzales J, de la Bellacasa JP, Ramirez of Klebsiella infection resistant to late-generation cephalosporins. Ann Intern
J, et al. Validation of different techniques for the diagnosis of ventilator- Med. 1993;119:353-8.
associated pneumonia. Comparison with immediate postmortem pulmonary 36. Tenover FC, Hughes JM. The challenges of emerging infectious diseases.
biopsy. Am J Respir Crit Care Med. 1994;149:324-31. Development and spread of multiply-resistant bacterial pathogens. JAMA.
9. Marquette CH, Copin MC, Wallet F, Neviere R, Saulnier F, Mathieu D, 1996;275:300-4.
et al. Diagnostic tests for pneumonia in ventilated patients: prospective eval- 37. Pittet D, Harbarth S. What techniques for diagnosis of ventilator-associated
uation of diagnostic accuracy using histology as a diagnostic gold standard. pneumonia? Lancet. 1998;352:83-4.

630 18 April 2000 • Annals of Internal Medicine • Volume 132 • Number 8