American Journal of Respiratory and Critical Care Medicine

To determine risk factors for ventilator-associated pneumonia (VAP) caused by potentially drug-resistant bacteria such as methicillin-resistant Staphylococcus aureus, Pseudomonas aeruginosa, Acinetobacter baumannii, and/or Stenotrophomonas maltophilia, 135 consecutive episodes of VAP observed in a single ICU over a 25-mo period were prospectively studied. For all patients, VAP was diagnosed based on results of bronchoscopic protected specimen brush ( ⩾ 103 cfu/ml) and bronchoalveolar lavage ( ⩾ 104 cfu/ml) specimens. Seventy-seven episodes were caused by “potentially resistant” bacteria and 58 episodes were caused by “other” organisms. According to logistic regression analysis, three variables among potential factors remained significant: duration of mechanical ventilation (MV) ⩾ 7 d (odds ratio [OR] = 6.0), prior antibiotic use (OR = 13.5), and prior use of broad-spectrum drugs (third-generation cephalosporin, fluoroquinolone, and/or imipenem) (OR = 4.1). Distribution of the 245 causative bacteria was analyzed according to four groups defined by prior duration of MV ( < 7 or ⩾ 7 d) and prior use or lack of use (within 15 d) of antibiotics. Although 22 episodes of early-onset VAP in patients receiving no prior antibiotics were caused by antibiotic-susceptible bacteria, 84 episodes of late-onset VAP in patients receiving prior antibiotics were mainly caused by potentially resistant bacteria. Differences in the potential efficacies (ranging from 100% to 11%) against microorganisms of 15 antimicrobial regimens were studied according to classification into these four groups. These findings may provide a more rational basis for selecting the initial therapy of patients suspected of having VAP.

The incidence of nosocomial pneumonia in ventilated patients is high, ranging from 7% to more than 40% (1). Such nosocomially acquired infections prolong hospital stay and contribute to ICU patient mortality (2-4). Despite extensive clinical experience with this disease, several sources of controversies still exist concerning epidemiologic patterns and optimal management strategies (5). Accurate data on etiologic agents and the epidemiology of ventilator-associated pneumonia (VAP) are limited by the lack of a “gold standard” for diagnosis. However, regardless of the diagnostic technique used, recent reports emphasize dramatic variations in the distributions of pathogens and drug-resistance patterns (5-8). Because early and appropriate antimicrobial therapy is an important goal in the setting of life-threatening infections (9-12), taking into account factors that modulate bacterial ecology and the susceptibility of causative organisms is crucial for optimal management.

In light of our well documented experience in the diagnosis of VAP (13-15) and our concern in the face of the increasing frequency of nosocomial pulmonary infections caused by multiresistant bacteria, we conducted a prospective study to better define risk factors linked to the emergence of a particular group of drug-resistant organisms: methicillin-resistant Staphylococcus aureus (MRSA), Pseudomonas aeruginosa, Acinetobacter baumannii, and Stenotrophomonas maltophilia. Using the two main risk factors identified by multivariate analysis, we were able to define four groups using duration of mechanical ventilation (MV) and prior use of antibiotic therapy as clinical parameters to classify each VAP episode. The distribution of causative organisms was analyzed according to this classification and the study was completed by studying the in vitro activities of 15 antimicrobial regimens in these four main groups of VAP episodes.

Study Population

Patients evaluated in this study were recruited from those hospitalized in one of two medical ICUs at a 1,300-bed teaching hospital, which serves as both a referral center and a first-line institution. It is a 17-bed ICU, only remarkable because of its specific recruitment of patients with multiple organ failure after cardiac surgery, representing a total of 32% of overall admissions, and a “shortage” of immunocompromised patients, who represented only 5% of the cases. The study period extended from May 1993 to June 1995.

All patients hospitalized and mechanically ventilated for more than 48 h were eligible for the study if pneumonia was suspected on clinical criteria and the patient's clinical status permitted flexible fiberoptic bronchoscopy. No patients were included whose antibiotic strategy had been changed during the 3 d preceding the diagnostic procedure. No systemic antibiotic regimen for nosocomial pneumonia prophylaxis or selective decontamination of the digestive tract was prescribed during the study period. No formal cycling of antibiotic regimens was used to prevent the emergence of resistant germs. All patients received sucralfate as the only prophylaxis against stress ulcers. In all patients a heat-moisture exchanger (Filter Dar Hygroster; Mallinckcrodt Medical, Mirandola, Italy) was positioned between the Y-piece and the patient. The heat-moisture exchangers were changed every 48 h. The oropharyngeal cavity was carefully cleaned four times daily with an antiseptic solution. All patients were kept in the semirecumbent position during most of their period of MV, except if not possible. To prevent nosocomial infection, effective surveillance, staff education, identification of high-risk patient, proper isolation techniques, and practices such as handwashing and the use of gloves for each patient-contact were the other systematic measures. Four hundred ninety-nine patients were eligible for the study.

Diagnosis of VAP

VAP was defined as any lower respiratory tract infection that developed after 2 d of MV. The criteria for clinical suspicion of pneumonia were as follows: presence of a new or persistent lung opacity on chest radiographs plus two of the following items: (1) fever > 38.3° C or hypothermia < 36° C, (2) WBC count > 10,000/mm3 or < 5,000/mm3, and (3) purulent endotracheal aspirate. Every patient suspected of having pneumonia underwent fiberoptic bronchoscopy to obtain samples by means of a protected specimen brush (PSB) and bronchoalveolar lavage (BAL). Specimens were collected and processed according to procedures extensively described elsewhere (15, 16). VAP was diagnosed based on results of PSB and BAL according to the following “significant” thresholds: PSB cultures yielding ⩾ 103 cfu/ml, ⩾ 5% of the recovered cells containing intracellular bacteria (ICB) upon direct examination of BAL fluid, and/or BAL cultures yielding ⩾ 104 cfu/ml. Bacterial identification and susceptibility tests were performed using standard methods. VAP was considered to be caused by “potentially resistant” bacteria when MRSA, P. aeruginosa, A. baumannii, and/or S. maltophilia grew at significant concentrations from PSB and/ or BAL specimens. For the purpose of this study, only the first episode of nosocomial pneumonia was taken into account.

Factors Associated with VAP Caused by Potentially Resistant Bacteria

To determine risk factors, the group “potentially resistant” bacteria was compared with the group “other” bacteria, which included VAP episodes caused by microorganisms other than MRSA, P. aeruginosa, A. baumannii, or S. maltophilia. VAP episodes that were caused by a “potentially resistant” pathogen plus “other” bacteria were classified as “potentially resistant” episodes. To perform this comparison, the following information on each patient was collected:

Parameters recorded within the first 24 h after admission. Age; sex; severity of the underlying medical condition stratified according to the criteria of McCabe and Jackson (17) as fatal, ultimately fatal, or not fatal; referral from another ICU; indication for MV using the classification described by Zwillich and coworkers (18); presence at admission of pneumonia, “sepsis,” or “sepsis with septic shock” according to the definitions of Bone and coworkers (19); presence of chronic obstructive pulmonary disease (COPD) and presence of acute respiratory distress syndrome (ARDS); simplified acute physiologic score II (20); acute physiology and chronic health evaluation (APACHE II) score (21); and total number of failed organs according to the definitions established in the organ dysfunctions and/or infection (ODIN) model (22).

Parameters recorded at the time of suspected nosocomial pneumonia. Temperature; WBC count; PaO2 /Fi O2 ; radiologic score according to the definition previously used by Fagon and coworkers (14); duration of MV before nosocomial pneumonia, recorded as a continuous and also a dichotomous variable with a cutoff at Day 7; total number of cells recovered by BAL; percentage of BAL cells containing ICB; number of causative organisms; and presence or absence of any antimicrobial agents for more than 24 h during the 15 d preceding the episode. Moreover, usage of each of the five following antibiotic classes during the 15 d preceding the event were recorded for each patient; imipenem, third-generation cephalosporin, aminoglycoside, fluoroquinolone, and/or “other” antibiotics (defined by exclusion). Broad-spectrum drugs (imipenem, third-generation cephalosporin and fluoroquinolone) were regrouped to constitute a new variable. Finally, the number of antibiotic classes administered during the 15 d prior to fiberoptic bronchoscopy was also recorded.

Definition of Four Groups of Patients

Four groups of patients were defined using prior duration of mechanical ventilation (< 7 or ⩾ 7 d) and presence or absence of antibiotics during the 15 d preceding the episode. Group 1 included patients ventilated for < 7 d without any prior antibiotic therapy, Group 2 included patients ventilated for < 7 d who had received at least one antibiotic within the previous 15 d, Group 3 included patients ventilated for ⩾ 7 d who had not received any antibiotic during the 15 preceding d, and Group 4 included patients ventilated for ⩾ 7 d who had received antibiotic therapy within the previous 15 d.

Microorganism Susceptibility

Among all the antibiotics tested against each bacterium cultured at significant level, 15 were considered for comparison among groups: amoxicillin, amoxicillin-clavulanic acid, ticarcillin, ticarcillin-clavulanic acid, piperacillin, piperacillin-tazobactam, cefamandole, cefotaxime, ceftazidime, aztreonam, imipenem, gentamicin, amikacin, vancomycin, and ciprofloxacin. Microorganism susceptibilities were determined using the criteria established by the “Comité National de l'Antibiothérapie,” the official committee in our country responsible for this classification. Results are expressed as percentages of susceptible bacteria. Because a high percentage of VAP episodes was polymicrobial, the potential efficacy of each antimicrobial agent was also calculated using susceptibility tests of all bacteria recovered at a significant level during each distinct episode.

Statistical Analyses

Potential risk factors, which are all the parameters listed in the above section, were subjected to univariate analyses in order to identify those significantly associated with VAP caused by drug-resistant bacteria. Continuous variables were compared using Student's t test; when not appropriate, the Mann-Whitney U-test was used. Chi-square statistics were used for categorical variables; when not appropriate, Fisher's exact test was used. Differences between groups were considered to be significant for variables yielding a p value ⩽ 0.05. Only those variables attaining an alpha value of 0.05 were included in the multiple logistic regression analysis model, and stepwise variable selection was performed to define the best-fitting model. Each variable used in the regression analysis was binary; therefore, continuous variables were recoded using the median value of the distribution in the absence of a clinically defined cutoff point. No interactions were tested between the variables. The discriminant ability of the model was assessed using the area under the receiver operating characteristic (ROC) curve (AUC). AUCs were compared using the test described by Hanley and McNeil (23). The Kappa test was used to compare multiple categorical measurements such as susceptibility to antimicrobial agents performed on the same variable. All analyses were performed using the StatView software package (version 4.11; Abacus Concepts, Berkeley, CA); the SAS software package (version 6.10; SAS Institute, Cary, NC) was used to perform the multiple logistic regression analyses.

Risk Factors for Developing VAP Caused by Potentially Resistant Bacteria

One hundred thirty-five patients (27.1%) developed at least one episode of nosocomial pneumonia. MRSA, P. aeruginosa, A. baumannii, or S. maltophilia (“potentially resistant” microorganisms) were found in 77 episodes (57.0%), “other” bacteria were involved in 58 episodes (43%). Epidemiologic characteristics of VAP caused by “potentially resistant” organisms and VAP caused by “other” bacteria are shown in Tables 1 2 3. According to univariate analysis, factors associated with the development of nosocomial pneumonia caused by “potentially resistant” microorganisms were sepsis present at admission and, at the time of VAP, duration of MV, severity of the radiologic score, percentage of infected BAL cells, and prior antibiotic therapy (Table ). When these variables were introduced into a logistic regression model, only the duration of MV before the occurrence of nosocomial pneumonia, prior antibiotic use, and prior use of broad-spectrum drugs (i.e., imipenem, third-generation cephalosporin, and/or fluoroquinolone) were found to be significant discriminant factors. The discriminant value of the model as determined by AUC was 0.89 (Table 4). When the variable “prior use of broad-spectrum antibiotics” was deleted from the model, the discriminant value was slightly lower (average c statistic, 0.86), but not significantly so (p = 0.27). Considering the duration of MV as a continuous instead of a dichotomous variable or using a cutoff of 5 d did not improve the accuracy of the model.


Characteristic“Potentially Resistant” Bacteria*(n = 77 )“Other” Bacteria(n = 58)
Age, yr, mean ± SD62.8 ± 13.262.4 ± 13.9
Sex, male, n (%)50 (64.9)42 (72.4)
Underlying medical condition, ultimately or rapidly fatal, n (%)23 (29.9)23 (39.7)
Referral from another ICU, n (%)20 (26)20 (34.5)
Indication for MV, n (%)
 Exacerbation of COPD4 (5.2)4 (6.9)
 Other pulmonary disease18 (23.4)12 (20.7)
 Postoperative respiratory failure51 (66.2)33 (56.9)
 Drug overdose00
 Neurologic emergency4 (5.2)4 (6.9)
 Miscellaneous05 (8.6)
Pneumonia, n (%) 8 (10.4)4 (6.9)
Sepsis, n (%)39 (50.6)18 (31.0)
Septic shock, n (%) 29 (37.7)14 (24.1)
COPD, n (%)14 (18.2)10 (17.2)
ARDS, n (%)39 (50.6)16 (27.6)
SAPS II, mean ± SD44.9 ± 13.245.2 ± 13.5
APACHE II score, mean ± SD22 ± 6.622 ± 6.4
Total number of dysfunctional organs, mean ± SD§ 2.7 ± 1.12.6 ± 1.0

*Corresponds to episodes in which at least one of the following microorganisms was involved: MRSA, P. aeruginosa, A. baumannii, and/ or S. maltophilia.

Corresponds to episodes in which no “multiresistant” bacteria were recovered at a significant level.

p = 0.02

§Organ dysfunctions are defined according to the ODIN model (22).


Variable“Potentially Resistant” Bacteria (n = 77 )“Other” Bacteria (n = 58)p Value
Temperature, °C, mean ± SD38.4 ± 1.338.4 ± 1.0NS
WBC × 103/ml, mean ± SD17.0 ± 8.815.6 ± 9.5NS
PaO2 /Fi O2 , mm Hg, mean ± SD205 ± 81197 ± 90NS
Radiologic score, mean ± SD6.2 ± 2.35.2 ± 2.6     0.02
Duration of MV before VAP onset, d, mean ± SD23 ± 179 ± 2< 0.0001
VAP episode occurring after ⩾ 7 d of MV, n (%)73 (94.8)28 (48.3)< 0.0001
Total number of BAL cells × 103/ml, mean ± SD1,729 ± 2,7421,612 ± 1,680NS
Percentage of BAL cells containing ICB, mean ± SD11.1 ± 15.217.9 ± 18.6     0.004
Number of causative organisms, mean ± SD1.9 ± 1.01.9 ± 0.9NS


Antibiotic History“Potentially Resistant” Bacteria (n = 77 )“Other” Bacteria (n = 58)Odds Ratiop Value
Prior therapy, n (%)74 (96.1)22 (37.9)40.4< 0.0001
Imipenem, n (%)14 (18.2)1 (1.7)12.7     0.002
Third-generation cephalosporin, n (%)34 (44.2)3 (5.2)14.5< 0.0001
Aminoglycoside, n (%)45 (58.4) 6 (10.3)12.2< 0.0001
Fluoroquinolone, n (%)16 (20.7)1 (1.7)14.9     0.001
Others, n (%)57 (74.0)20 (34.5) 5.4< 0.0001
Number of antibiotic classes previously given,  mean ± SD* 2.16 ± 1.000.54 ± 0.57 6.8< 0.0001
Broad-spectrum drug, n (%) 52 (67.5)5 (8.6)22.2< 0.0001

*According to the above classification, maximum value = 5.

Any of the following antibiotic classes: imipenem, third-generation cephalosporins, and/or fluoroquinolone.


VariableOdds Ratio95% Confidence Intervalp Value
Duration of MV before VAP episode ⩾ 7 d (yes/no) 6.011.6–23.10.009
Prior antibiotic use (yes/no)13.463.3–55.00.0003
Broad-spectrum antibiotics (yes/no) 4.121.2–14.20.025

*Discriminant value (AUC) = 0.89.

Distribution of Microorganisms Responsible for VAP

A total of 245 bacteria was cultured at a significant concentration. Seventy-four (54.8%) of the 135 episodes were polymicrobial. Cocci alone and bacilli alone accounted for 31% and 41.5% of the episodes, respectively, whereas cocci and bacilli in combination were causative agents in 37 episodes (27.4%). The most frequently isolated organisms were S. aureus (21.3%), Enterobacteriaceae (17.9%), P. aeruginosa (15.9%), nonpneumococcal Streptococcus species (13.5%), and A. baumannii (9%) (Table 5). Streptococcus pneumoniae accounted for less than 2% of the total number of bacteria in this population of patients mechanically ventilated for more than 48 h. No infections were caused by nonfermenting gram-negative bacilli and/or MRSA before Day 7 of MV in patients who had not recently received antimicrobial treatment (Table 6).


Microorganisms* n (%)
Bacilli128 (52.2)
Pseudomonas aeruginosa  39 (15.9)
Acinetobacter baumannii 22 (9)
Stenotrophomonas maltophilia  6 (2.4)
Klebsiella species 9 (3.7)
Escherichia coli  8 (3.3)
Proteus species 7 (2.9)
Enterobacter species 5 (2.0)
Morganella species 4 (1.6)
Serratia species 4 (1.6)
Hafnia species 4 (1.6)
Citrobacter diversius  3 (1.2)
Hemophilus species15 (6.1)
Corynebacterium species 2 (0.8)
Cocci111 (45.3)
 MRSA 32 (13.1)
 MSSA20 (8.2)
Streptococcus pneumoniae  3 (1.2)
Streptococcus species 33 (13.5)
Enterococcus species 5 (2.0)
Neisseria species14 (5.7)
 CNS 4 (1.6)
Anaerobic pathogens 6 (2.4)
Total245 (100)

Definition of abbreviations: MSSA = methicillin-sensitive Staphylococcus aureus; CNS = coagulase-negative staphylococci.

*In nine episodes, fungal agents (Candida sp.) were concomitantly isolated at concentrations ⩾ 103 cfu/ml from PSB or ⩾ 104 cfu/ml from BAL fluid.


OrganismsGroup 1 (n = 22) MV < 7 ABT = noGroup 2 (n = 12) MV < 7 ABT = yesGroup 3 (n = 17) MV ⩾ 7 ABT = noGroup 4 (n = 84) MV ⩾ 7 ABT = yes
Multiresistant bacteria0*  6 (30) 4 (12.5) 89 (58.6)
P. aeruginosa 0 4 (20)2 (6.3)33 (21.7)
A. baumannii 01 (5)1 (3.1)20 (13.2)
S. maltophilia 0006 (3.9)
 MRSA01 (5)1 (3.1)30 (19.7)
Other bacteria41 (100)14 (70)28 (87.5)63 (41.4)
 Enterobacteriaceae10 (24.4) 4 (20) 7 (21.9)23 (15.1)
Hemophilus spp. 8 (19.5) 2 (10)1 (3.1)4 (2.6)
 MSSA 6 (14.6)0 7 (21.9)7 (4.6)
S. pneumoniae 3 (7.3)000
 Other streptococci 7 (17.1) 5 (25) 7 (21.9)14 (9.2)
Neisseria spp. 5 (12.2) 2 (10) 4 (12.5)3 (2)
 Other pathogens2 (4.9)1 (5)2 (6.3)12 (7.9)
Total number of bacteria41 (100)20 (100)32 (100)152 (100)

*p < 0.02 versus Groups 2, 3, or 4.

p < 0.0001 versus Group 4.

On the basis of this fact and on the main results of the multivariate analysis, prior use of antibiotic and MV duration ⩾ 7 d were used to define four groups: Group 1 included 22 patients ventilated for < 7 d without receiving any prior antibiotic therapy; Group 2 included 12 patients ventilated for < 7 d who had received at least one antimicrobial agent within the previous 15 d (no patient in this group had received broad-spectrum antibiotics); Group 3 included 17 patients ventilated for ⩾ 7 d who had not received any antibiotic therapy during the preceding 15 d; Group 4 included 84 patients ventilated for ⩾ 7 d who had received antibiotic therapy within the previous 15 d. Seventy percent of the patients in Group 4 had received at least one broad-spectrum antibiotic.

The distribution of causative bacteria among these four groups is detailed in Table . The distribution of “potentially resistant” bacteria as compared with “other” bacteria differed between groups (p < 0.0001). Group 1 differed significantly from every other group (p < 0.02), and Group 3 from Group 4 (p < 0.0001). Early-onset episodes in patients who had not received previous antibiotic therapy (Group 1) were caused primarily by Enterobacteriaceae, Hemophilus spp., MSSA, and/ or streptococci. Conversely, in patients who were ventilated ⩾ 7 d and who had received previous antibiotic therapy (Group 4), VAP was mainly caused by multiresistant bacteria such as P. aeruginosa, A. baumannii, S. maltophilia, or MRSA. For the two intermediate groups, early-onset episodes with previous antibiotic therapy (Group 2) and late-onset episodes without previous antibiotic therapy (Group 3), a mixed distribution of pathogens was observed. S. pneumoniae was rarely recovered and only in Group 1. Hemophilus spp. were more frequently found in episodes corresponding to patients ventilated for < 7 d (10 of 34 episodes in Groups 1 and 2 versus five of 101 episodes in Groups 3 and 4) (p = 0.0003). P. aeruginosa was present in 38.5% of the episodes occurring in patients who had received antibiotics versus 5% of the episodes that developed in patients who had not received any previous antibiotic therapy (p < 0.01). S. maltophilia and A. baumannii were quasi-exclusively found in Group 4.

Microorganism Susceptibility According to Duration of MV and Prior Use of Antibiotics

Overall, the susceptibilities of causative organisms to various antibiotics ranged from 34 to 76%. The potential efficacy of each antimicrobial regimen differed significantly between groups, except for amikacin (Table 7). Analysis of Group 1 showed that amoxicillin-clavulanic acid, cefamandole, piperacillin-tazobactam, cefotaxime, and ceftazidime were active against 90 to 100% of the etiologic agents. In this group, all isolates except one organism, Hemophilus parainfluenzae, which was recovered from a chronically ill patient who had been hospitalized frequently before referral to our ICU, were susceptible to imipenem.


RegimenGroup 1 (22/41)*MV < 7 d ABT = noGroup 2 (12/20)*MV < 7 d ABT = yesGroup 3 (17/32)*MV ⩾ 7 d ABT = noGroup 4 (84/152)*MV ⩾ 7 d ABT = yesp Value
Amoxicillin 52595323< 0.0001
Amoxicillin-clavulanic acid 90607232< 0.0001
Ticarcillin 69805643     0.001
Ticarcillin-clavulanic acid 85826350     0.0001
Piperacillin 73766644     0.0006
Piperacillin-tazobactam100859458< 0.0001
Cefamandole 93657228< 0.0001
Cefotaxime100708134< 0.0001
Ceftazidime 98808850< 0.0001
Aztreonam 55473428     0.007
Imipenem 98959464< 0.0001
Gentamicin 62505333     0.004
Amikacin 63605542     0.06
Ciprofloxacin 69606338     0.0006

*Number of episodes/number of pathogens.

As expected, the levels of susceptibility of bacteria isolated from Group 4 patients were markedly reduced, and only four antimicrobial regimens showed an efficacy ⩾ 50%: ticarcillin-clavulanic acid, ceftazidime, piperacillin-tazobactam, and imipenem.

Detailed results of gram-negative bacilli susceptibility are given in Table 8. For Group 1, only four strains, all Enterobacteriaceae, were resistant to amoxicillin-clavulanic acid (Enterobacter cloacae, n = 2; Hafnia spp., n = 1; and Serratia marcescens, n = 1). For Groups 2 and 3, other than ceftazidime and imipenem, ticarcillin, piperacillin-tazobactam, aztreonam, aminoglycosides, and ciprofloxacin had potential efficacies superior or close to 75%. For Group 4, imipenem was the only drug with an antimicrobial activity equal to 75%, even though no extended-spectrum beta-lactamase-producing Klebsiella pneumoniae were isolated during the study period. Amikacin was more potent than gentamicin (p < 0.001). Subanalysis of gram-negative bacillus susceptibility in Group 4 according to prior exposure to broad-spectrum drugs, which was the case for 63 of 90 bacteria, showed significantly lower efficacy for piperacillin-tazobactam (58% versus 83%, p = 0.04), cefotaxime (19% versus 50%, p = 0.006), aztreonam (38% versus 74%, p = 0.007), and ciprofloxacin (47% versus 75%, p = 0.03), and no statistically significant difference was found for the 10 other antimicrobial regimens tested. Only imipenem had a potential efficacy > 70% (71%) against gram-negative bacilli in the subgroup of patients who had received prior broad-spectrum antibiotics. In contrast, seven antibiotic regimens given to patients in the subgroup who had received antibiotics but no broad-spectrum drugs had a potential efficacy ⩾ 70%: ticarcillin-clavulanic acid (70%), piperacillin-tazobactam (85%), ceftazidime (75%), aztreonam (74%), amikacin (75%), ciprofloxacin (75%), and imipenem (83%).


RegimenGroup 1 (22/19)*MV < 7 d ABT = no Group 2 (12/12)*MV < 7 d ABT = yesGroup 3 (17/11)*MV ⩾ 7 d ABT = noGroup 4 (84/90)*MV ⩾ 7 d ABT = yesp Value
Amoxicillin 33 252711     0.07
Amoxicillin-clavulanic acid 78 362719< 0.0001
Ticarcillin 67 737251     0.23
Ticarcillin-clavulanic acid100 757265     0.02
Piperacillin 76 626449     0.19
Piperacillin-tazobactam100 829165     0.008
Cefamandole 89 454519< 0.0001
Cefotaxime100 557328< 0.0001
Ceftazidime100 829159     0.001
Aztreonam100 809149< 0.0001
Imipenem 941009175     0.054
Gentamicin 83 737349     0.02
Amikacin 83 918065     0.15
Ciprofloxacin 94 918255     0.001

*Number of episodes/number of pathogens.

Antibiotic therapy.

Information concerning the antimicrobial resistance of gram-positive cocci is given in Table 9. Each of the 16 isolates from Group 1 was susceptible to amoxicillin-clavulanic acid, piperacillin-tazobactam, cefamandole and cefotaxime, ceftazidime, and of course imipenem and vancomycin. For Groups 2 and 3, the percentages of species susceptible to all these antimicrobials were similar and > 80%. For Group 4, the potential efficacy of all antibiotics listed in the table, except vancomycin was < 45%, primarily because of the presence of MRSA and enterococci in these episodes. Vancomycin was the only drug active against all gram-positive cocci infections.


RegimenGroup 1 (22/16)*MV < 7 d ABT = noGroup 2 (12/6)*MV < 7 d ABT = yesGroup 3 (17/17)*MV ⩾ 7 d ABT = noGroup 4 (84/58)*MV ⩾ 7 d ABT = yesp Value
Amoxicillin 63 83 59 29     0.004
Amoxicillin-clavulanic acid100 83 94 43< 0.0001
Ticarcillin 63 83 47 29     0.01
Ticarcillin-clavulanic acid 63 83 47 29     0.01
Piperacillin 63 83 59 29     0.004
Piperacillin-tazobactam100 83 94 42< 0.0001
Cefamandole100 83 83 37< 0.0001
Cefotaxime100 83 83 37< 0.0001
Ceftazidime100 83 82 37< 0.0001
Imipenem100 83 94 44< 0.0001
Gentamicin 37  0 41 14     0.02
Ciprofloxacin 37  0 41 14     0.02

*Number of episodes/number of pathogens.

Because empiric therapy of pneumonia should cover all pathogens recovered at a significant level, Table 10 lists the susceptibility rates to antimicrobial regimens of all the microorganisms recovered during each episode. For Group 1, the potential activity of piperacillin-tazobactam and cefotaxime was 100%. For Groups 2 and 3, the highest susceptibility percentages were obtained with imipenem, piperacillin-tazobactam, and ceftazidime. Adjunction of either an aminoglycoside or ciprofloxacin to beta-lactams extended only poorly, if at all, the spectrum of activity when the initial score was in the range of 70% or higher. For Group 4, no single regimen had a potential activity superior to 40%. In this latter situation, combination therapy with a glycopeptide was mandatory to increase the potential efficacy of the antimicrobial treatment to above 70%. Combining beta-lactams with one of the following drugs—gentamicin, amikacin, or ciprofloxacin—only slightly improved the potential activity of treatment. Examples of scores obtained using such combination therapy to treat patients assigned to Group 4 are shown in Figure 1. The most potent regimen combined imipenem plus amikacin plus vancomycin and generated a potential efficacy of 88%.


RegimenGroup 1 (n = 22) MV < 7 d ABT = noGroup 2 (n = 12) MV < 7 d ABT = yesGroup 3 (n = 17) MV ⩾ 7 d ABT = noGroup 4 (n = 84) MV ⩾ 7 d ABT = yesp Value
Amoxicillin 323325 2< 0.0001
Amoxicillin-clavulanic acid 82335312< 0.0001
Ticarcillin 45674120     0.003
Ticarcillin-clavulanic acid 76704131     0.0005
Piperacillin 52603823     0.01
Piperacillin-tazobactam100758835< 0.0001
Cefamandole 86425312< 0.0001
Cefotaxime100507115< 0.0001
Ceftazidime 96677732< 0.0001
Aztreonam 41272918     0.14
Imipenem 96928839< 0.0001
Gentamicin 45335920     0.005
Amikacin 45425925     0.03
Ciprofloxacin 50425924     0.01

To provide a comprehensive description of microorganisms responsible for VAP and to determine risks factors leading to the emergence of potentially resistant pathogens, we undertook a prospective study on a series of consecutive ICU patients who required MV for > 48 h, using strict criteria to define pneumonia and applying multivariate analysis to simultaneously take into account variables known to be potentially associated with selection of drug-resistant bacteria. Our results indicate that the incidence of VAP caused by potentially drug-resistant bacteria was high (57%) and that three variables remained independently associated with these infections: duration of MV ⩾ 7 d (odds ratio [OR] = 6.0), prior antibiotic use (OR = 13.5), and prior use of broad-spectrum antibiotics (OR = 4.1). When the distribution of the causative bacteria isolated from the 135 episodes of VAP was analyzed according to the four groups of patients defined by prior duration of MV (< 7 or ⩾ 7 d) and presence or absence of antibiotics during the 15 d preceding the event, important differences were brought to light. Although early-onset pneumonia in patients who had not received prior antimicrobial treatment was mainly caused by sensitive Enterobacteriaceae, Hemophilus spp., MSSA, and S. pneumoniae, late-onset pneumonia in patients who had recently received antimicrobial treatment was mostly caused by potentially resistant pathogens such as P. aeruginosa, A. baumannii, S. maltophilia, and MRSA. As a result, microorganism susceptibility to various antimicrobial agents differed widely from one group of patients to another. Although amoxicillin-clavulanic acid or a second-generation cephalosporin was active against ⩾ 90% of bacteria recovered from patients with early-onset pneumonia who had not received prior antimicrobial treatment, only four antimicrobial agents—ticarcillin-clavulanic acid, ceftazidine, piperacillin-tazobactam, and imipenem— showed potential efficacy slightly superior to 50% in patients who had undergone prolonged MV and received recent antimicrobial treatment, thereby emphasizing that monotherapy with these regimens would not have assured adequate coverage of all putative pathogens in this setting.

The importance of potentially resistant microorganisms, as defined in this report, as a cause of VAP has been documented previously (6, 7). Regardless of the bacteriologic method used to define the precise etiology of pneumonia, several studies have reported a high rate of nosocomial pneumonia caused by P. aeruginosa and MRSA (8, 9, 24). The especially high incidence (57%) of potentially resistant bacteria observed in our study population may, however, reflect a particular situation since our ICU accumulates risk factors for an increased rate of infection caused by multiresistant pathogens, despite the systemic use of invasive diagnostic techniques to avoid indiscriminate antibiotic usage in patients clinically suspect of having developed VAP. First, the prevalence of MRSA is particularly high in France; second, our ICU is located in a very large university hospital with more than 1,200 beds; third, nearly 30% of our patients were referred to our center from another ICU, very often after a prolonged and complicated course; fourth, our unit is largely specialized in the care of critically ill patients with acute infectious diseases who require treatment with antimicrobial agents, mostly broad-spectrum antibiotics, for their primary admission illness, even if we try to use a very strict antibiotic policy; and fifth, the percentage of patients who need prolonged (> 7 d) MV is particularly high in our unit (29%), a factor directly associated with infection caused by potentially resistant pathogens, as demonstrated in our study.

The findings obtained in this study using multivariate analysis confirmed the strong relationship, previously identified by means of univariate analysis, between the occurrence of infection caused by potentially resistant pathogens and time of onset of pneumonia and previous antibiotic therapy (25-28). In the study by Prod'hom and coworkers (25), early-onset pneumonia, occurring during the first 4 d of MV, were due to S. aureus, S. pneumoniae, and H. influenzae alone or in combination in 54% of the cases. Gram-negative organisms were found in 66% of the late-onset pneumonia episodes. Interestingly, in most of these earlier investigations a cutoff point of 4 or 5 d was used to distinguish early-onset from late-onset pneumonia (25-28). Because in our study population no infections were caused by a nonfermenting gram-negative bacillus or a MRSA before Day 7 of MV in patients who had not recently received antimicrobial treatment, we chose to use this cutoff time to separate patients with early onset from those with late-onset pneumonia.

The role of recent antibiotic therapy has also been underscored by several investigators. In the study by Fagon and coworkers (24), the majority (89%) of patients who developed pneumonia caused by Pseudomonas or Acinetobacter species had been receiving antimicrobial therapy prior to the onset of pneumonia, whereas only 17% of the pneumonias occurring in patients without such antibiotic therapy were due to these drug-resistant organisms. In the study of S. aureus VAP conducted by Rello and coworkers (29), all patients with MRSA VAP had recently received antibiotics, compared with only 21% of those with MSSA episodes. The same investigators also evaluated the impact of recent antimicrobial therapy on the etiology of VAP in a large series of patients who developed VAP (30). Their most striking finding was that the rate of pneumonia caused by P. aeruginosa was significantly higher in patients who had received prior antimicrobial treatment, whereas the rate of pneumonia caused by gram-positive cocci or H. influenzae was significantly lower. In agreement with these studies, our data strongly support the observation that recent prior use of antibiotics is a key factor in selecting for potentially resistant microorganisms, with an OR = 13.5 (95% CI, 3.3 to 55.0). As previously documented in several studies, not only the presence or absence of antimicrobial treatment before the onset of pneumonia but also the specific use of broad-spectrum antimicrobial agents such as third-generation cephalosporins, imipenem, or fluoroquinolones played a major role in the emergence of multiresistant organisms (31-33). Two factors in our study support this latter finding: first, addition of “broad-spectrum drugs” to the logistic regression model slightly improved the predictive value of the model; and, second, in patients with late-onset pneumonia who had received prior antimicrobial treatment, the susceptibility of gram-negative bacilli was significantly reduced with respect to several potent antibiotics when patients had received one of these broad-spectrum drugs.

Some potentially important variables such as prior treatment with corticosteroids and/or H2-blockers, and nutritional status were not considered in our analysis, thus, their role in the acquisition of potentially resistant bacteria could not be ascertained from our study.

Because the two main risk factors selecting for potentially resistant pathogens—prior duration of MV and recent use of antibiotics—are easy to record routinely, it was tempting to apply them to classify patients with VAP into four clinical groups. Using this classification, differences in the distributions of etiologic agents were evident. Whereas patients in Group 1 (duration of MV < 7 d, no prior antibiotics) only developed infections caused by microorganisms that are usually antibiotic-sensitive such as S. pneumoniae, MSSA, H. influenzae, and Enterobacteriaceae, most patients included in Group 4 (duration of MV ⩾ 7 d, prior use of antibiotics) developed infections caused by difficult-to-treat pathogens, such as MRSA, P. aeruginosa, A. baumannii, and S. maltophilia. Interestingly, MRSA, A. baumannii, and S. maltophilia were practically never isolated from patients with VAP other than those in Group 4. In contrast, a few P. aeruginosa infections were observed in patients assigned to Group 2 (duration of MV < 7 d, prior use of antibiotics), underlining that this type of VAP may occur very early in the course of MV once a patient has received prior antibiotics.

As expected, these differences in the distributions of VAP pathogens had direct consequences on microorganism susceptibilities to various antimicrobial agents. Taking these data into account might therefore allow the definition of a more rational decision tree for selecting initial treatments for patients strongly suspected of having VAP, a step that would avoid resorting to broad-spectrum drug coverage in all cases. For example, monotherapy with a second-generation cephalosporin or a nonpseudomonal third-generation cephalosporin, or a combined regimen in which a beta-lactamase inhibitor, clavulanic acid, is added to amoxicillin appeared to be appropriate choices for most of our patients with early-onset pneumonia who had not received prior antimicrobial treatment. In contrast, for patients who had required previous prolonged MV and antimicrobial treatment, especially with a “broad-spectrum drug,” a triple combination therapy with a carbapenem such as imipenem, which in our ICU is currently the only antimicrobial agent constantly active against A. baumannii, plus an aminoglycoside or ciprofloxacin plus vancomycin was the only regimen that achieved acceptable coverage of causative pathogens before culture results were available (see Table and Figure 1). For patients with early-onset pneumonia who had received prior antibiotics and patients with late-onset pneumonia who had not received prior antibiotics or only non-broad-spectrum drugs, combination therapy with either an antipseudomonal cephalosporin or an antipseudomonal penicillin with a beta-lactamase inhibitor plus an aminoglycoside or ciprofloxacin but without vancomycin was adequate. Concerning the use of aminoglycosides or ciprofloxacin, it should be noted that the addition of either of these two drugs extended poorly, if at all, the spectrum of activity when the potential efficacy of the selected beta-lactam was > 70%. However, because these agents may provide an additive or even a synergistic activity against P. aeruginosa or other multiresistant pathogens when combined with a beta-lactam antibiotic and prevent the emergence of resistance during therapy, we think it is probably safer to use such a combination in patients with severe nosocomial pneumonia, at least during the first days of therapy, while waiting for results of the microbiologic pulmonary secretion cultures.

When these propositions based on our results in 135 consecutive episodes of VAP for which the causative microorganisms were documented using bronchoscopic methods are compared with the recently published American Thoracic Society (ATS) guidelines for empiric treatment of nosocomial pneumonia, many similarities are evident and only a few differences can be noted (5). First, in our approach, the cutoff point to distinguish between early- and late-onset pneumonias was delayed from 5 to 7 d, which could avoid resorting to broad-spectrum coverage before that date in patients who had not received antibiotics recently. Second, because in spite of a very high incidence of MRSA, we almost never observed infection caused by this microorganism before the seventh day of MV and only in patients who had previously received antibiotics, the use of vancomycin could be restricted to patients having these two risk factors. Finally, the decision tree applied in the ATS guidelines for patients with severe pneumonia defines three categories of episodes with only two tables of antibiotic recommendations. The category “early-onset without risk factors” is very similar to our Group 1, and therapeutic recommendations are the same. However, the two other groups of patients identified in the ATS guidelines (“late-onset without risk factors” and “onset at any time with risk factors”) are to be treated according to the same algorithm, based on a combination of aminoglycoside or ciprofloxacin plus one of the following antimicrobial agents, an antipseudomonal penicillin, a beta-lactam/beta-lactamase inhibitor combination, ceftazidine or cefoperazone, imipenem or aztreonam with or without vancomycin. As described above, our classification will lead to two different therapeutic strategies for patients not included in Group 1, restricting the use of imipenem to patients with late-onset pneumonia who had previously received broad-spectrum antibiotics. Because the numbers of patients included in Groups 2 and 3 in our series were small, the exact impact of this new classification scheme on antibiotic therapy of patients with suspected VAP remains to be evaluated.

At least three limitations indicate that our data may not be widely applicable. First, our study was conducted in a single unit. Second, we used the duration of MV as a clinical parameter to categorize each episode of VAP. This definition does not preclude the possibility that patients may have a long stay in the hospital prior to the beginning of MV. In our recruitment, this event is uncommon, but this clinical situation represents a potential limitation of our classification scheme. This is exemplified by one of our patients hospitalized frequently before referral to our ICU who developed early-onset VAP caused by a multiresistant strain of Hemophilus parainfluenzae, although this patient had not received recent antibiotic therapy. Third, for the multiple reasons detailed above, we are confronted in our ICU with a high rate of nosocomial infection caused by potentially drug-resistant bacteria. However, many ICUs in Europe and North America are in fact also facing the same problem (6-8, 29-33).

In summary, the findings of this study indicate that the duration of MV prior to the onset of VAP and the recent (within 15 d) use of antibiotics are two key factors favoring the emergence of potentially resistant organisms responsible for VAP. Although early-onset VAP in patients with no prior antibiotics are caused by antibiotic-susceptible bacteria, late-onset VAP in patients who had recently received antibiotics are mainly caused by multiresistant bacteria such as P. aeruginosa, A. baumannii, S. maltophilia, and MRSA. These data, if they are confirmed in other ICU patients, may provide a more rational basis for selecting initial therapy for VAP patients before culture results are available.

The writers thank Mrs. C. Brun and A. Failin for assistance in the preparation of the manuscript.

1. Chastre, J., and J. Y. Fagon. 1994. Pneumonia in the ventilator-dependent patient. In M. J. Tobin, editor. Principles and Practice of Mechanical Ventilation. McGraw-Hill, New York. 857–890.
2. Craig C. P., Connelly S.Effect of intensive care unit nosocomial pneumonia on duration of stay and mortality. Am. J. Infect. Control121984233238
3. Fagon J. Y., Chastre J., Hance A. J., Montravers P., Novara A., Gibert C.Nosocomial pneumonia in ventilated patients. Am. J. Med.941993281288
4. Fagon J. Y., Chastre J., Vuagnat A., Trouillet J. L., Novara A., Gibert C.Nosocomial pneumonia and mortality among patients in intensive care units. J.A.M.A.2751996866869
5. Campbell G. D., Niederman M. S., Broughton W. A., Craven D. E., Fein A. M., Fink M. P., Gleeson K., Hornick D. B., Lynch J. P., Mandell L. A., Mason C. M., Torres A., Wunderink R. G.Hospital-acquired pneumonia in adults: diagnosis, assessment of severity, initial antimicrobial therapy, and preventive strategies. A consensus statement. Am. J. Respir. Crit. Care Med.153199617111725
6. Shaberg, D. R., D. H. Culver, and R. P. Gaynes. 1991. Major trends in the microbial etiology of nosocomial infection. Am. J. Med. 91(Suppl. 3B):72S–75S.
7. Emori T. G., Gaynes R. P.An overview of nosocomial infections, including the role of the microbiology laboratory. Clin. Microbiol. Rev.61993428442
8. Spencer R. C.Predominant pathogens found in the European Prevalence of Infection in Intensive Care Study. Eur. J. Clin. Microbiol. Infect. Dis.151996281285
9. Torres A., Aznar R., Gatell J. M., Jimenez P., Gonzalez J., Ferrer A., Celis R., Rodriguez-Roisin R.Incidence, risk and prognosis factors of nosocomial pneumonia in mechanically ventilated patients. Am. Rev. Respir. Dis.1421990523528
10. Celis R., Torres A., Gatell J. H., Almela M., Rodriguez-Roisin R., Augusti-Vidal A.Nosocomial pneumonia: a multivariate analysis of risk and prognosis. Chest931988318324
11. Kollef M. H.Ventilator-associated pneumonia: a multivariate analysis. J.A.M.A.270199319651970
12. Alvarez-Lerma F.the ICU-Acquired Pneumonia Study GroupModification of empiric antibiotic treatment in patients with pneumonia acquired in the ICU. Intensive Care Med.221996387394
13. Chastre J., Viau F., Brun P., Pierre J., Dauge M. C., Bouchama A., Akesbi A., Gibert C.Prospective evaluation of the protected specimen brush for the diagnosis of pulmonary infections in ventilated patients. Am. Rev. Respir. Dis.1301984924929
14. Fagon J. Y., Chastre J., Hance A. J., Guiguet M., Trouillet J. L., Domart Y., Pierre J., Gibert C.Detection of nosocomial lung infection in ventilated patients: use of a protected specimen brush and quantitative culture techniques in 147 patients. Am. Rev. Respir. Dis.1381988110116
15. Chastre J., Fagon J. Y., Bornet-Lecso M., Calvat S., Dombret M. C., Khani R. A., Basset F., Gibert C.Evaluation of bronchoscopic techniques for the diagnosis of nosocomial pneumonia. Am. J. Respir. Crit. Care Med.1521995231240
16. Chastre, J., J. Y. Fagon, and J. L. Trouillet. 1995. Diagnosis and treatment of nosocomial pneumonia in patients in intensive care units. Clin. Infect. Dis. 21(Suppl. 3):S226–S237.
17. McCabe W. R., Jackson G. G.Gram-negative bacteremia, I. Arch. Intern. Med.1101982847864
18. Zwillich C., Pierson D., Creagh C., Sutton S., Schatz E., Petty T.Complications of assisted ventilation: a prospective study of 354 consecutive episodes. Am. J. Med.571974161170
19. Bone R. C., Balk R. A., Cerra F. B., Dellinger R. P., Fein A. M., Knaus W. A., Schein R. M. H., Sibbald W. J.Definitions for sepsis and organ failure and guidelines for use of innovative therapies in sepsis. Chest101199216441655
20. Legall J. R., Lemeshow S., Saulnier F.A new simplified acute physiology score (SAPS II) based on a European/North American multicenter study. J.A.M.A.270199329572963
21. Knauss W. A., Drapper E. A., Wagner D. P., Zimmerman J. E.Apache II: a severity of disease classification system. Crit. Care Med.131985818829
22. Fagon J. Y., Chastre J., Novara A., Medioni P., Gibert C.Characterization of intensive care unit patients using a model based on the presence or absence of organ dysfunctions and/or infection: the ODIN model. Intensive Care Med.191993137144
23. Hanley J. A., McNeil B. J.A method of comparing the areas under receiver operating characteristic curves derived from the same cases. Radiology1481983839843
24. Fagon J. Y., Chastre J., Domart Y., Trouillet J. L., Pierre J., Darne C., Gibert C.Nosocomial pneumonia in patients receiving continuous mechanical ventilation: prospective analysis of 52 episodes with use of a protected specimen brush and quantitative culture techniques. Am. Rev. Respir. Dis.1391989887894
25. Prod'hom G., Leuenberger P., Koerfer J., Blum A., Chiolero R., Schaller M. D., Perret C., Spinnler O., Blondel J., Siegrist H., Saghafi L., Blanc D., Francioli P.Nosocomial pneumonia in mechanically ventilated patients receiving antiacid, ranitidine, or sucralfate as prophylaxis for stress ulcer. Ann. Intern. Med.1201994653662
26. Baker A. M., Meredith J. W., Haponik E. F.Pneumonia in intubated trauma patients: microbiology and outcomes. Am. J. Respir. Crit. Care Med.1531996343349
27. Kollef M. H., Silver P., Murphy D. M., Trouillon E.The effect of late-onset ventilator-associated pneumonia in determining patient mortality. Chest108199516551662
28. Chevret S., Hemmer M., Carlet J., Langer M.the European Cooperative Group on Nosocomial PneumoniaIncidence and risk factors of pneumonia acquired in intensive care units: results from a multicenter prospective study on 996 patients. Intensive Care Med.191993256264
29. Rello J., Torres A., Ricart M., Vallés J., Gonzalez D., Artigas A., Rodriguez-Roisin R.Ventilator-associated pneumonia by Staphylococcus aureus: comparison of methicillin-resistant and methicillin-sensitive episodes. Am. J. Respir. Crit. Care Med.150199415451549
30. Rello J., Ausina V., Ricart M., Castella J., Prats G.Impact of previous antimicrobial therapy on the etiology and outcome of ventilator associated pneumonia. Chest104199312301235
31. Jones, R. N. 1996. Impact of changing pathogens and antimicrobial susceptibility patterns in the treatment of serious infections in hospitalized patients. Am. J. Med. 100(Suppl. 6A):3S–12S.
32. Ballow C. H., Schentag J. J.Trends in antibiotic utilization and bacterial resistance. Diagn. Microbiol. Infect. Dis.15199237S42S
33. Jacobson K., Cohen S. H., Inciardi J. F., King J. H., Lippert W. E., Iglesias T., Vancouwenberghe C. J.The relationship between antecedent antibiotic use and resistance to extended-spectrum cephalosporins in group 1 β-lactamase-producing organisms. Clin. Infect. Dis.21199511071113
Correspondence and requests for reprints should be addressed to Dr. Jean-Louis Trouillet, Service de Réanimation Médicale, Hôpital Bichat, 46, rue Henri-Huchard, 75877 Paris Cedex, France.


No related items
American Journal of Respiratory and Critical Care Medicine

Click to see any corrections or updates and to confirm this is the authentic version of record