INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Keywords: ventilator-associated pneumonia; bronchoscopy; routine surveillance cultures; empiric antimicrobial treatment

Most investigators agree that inadequate initial treatment of nosocomial pneumonia in patients requiring intensive care unit (ICU) admission is an important determinant of hospital mortality and that clinical efforts aimed at rectifying this therapeutic failure could improve the outcomes of critically ill patients (1).

Several investigators have recommended routine surveillance cultures of ICU patients because they may be predictive of patients who are at high risk of invasive disease and, furthermore, should invasive disease develop, empiric therapy can be selected based on the predominant pathogens identified in these cultures (5). However, the accuracy of this approach for selecting initial antimicrobial treatment in ICU patients requiring new antibiotics for ventilator-associated pneumonia (VAP) has not yet been established.

To test the hypothesis that results of routine microbiologic cultures of specimens obtained before the onset of VAP in ICU patients may help to predict the causative microorganisms and thus to select effective initial antimicrobial therapy, we prospectively studied a large series of consecutive VAP episodes for which the causative microorganisms were determined using bronchoscopic techniques. Our specific goals were: (1) to compare the microorganisms responsible for VAP with those previously recovered; (2) to document the specific value of respiratory specimens obtained before onset of VAP, because this microbiologic information might be particularly useful for identifying the responsible organisms in the event of subsequent pneumonia; and (3) to determine the sensitivity and specificity of previous colonization or infection with potentially drug-resistant microorganisms, such as methicillin-resistant Staphylococcus aureus (MRSA), Pseudomonas aeruginosa, or Acinetobacter baumannii, to predict nosocomial pneumonia caused by these organisms.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients evaluated in this study were recruited from among those hospitalized in the 17-bed medical ICU of Bichat University Hospital, which serves as both a referral center and a first-line institution, receiving patients from the community, from several other wards and specialized ICUs in the hospital, and from ICUs of other hospitals. The study period was from February 1997 to June 1998. In accordance with French law, no informed consent was mandatory, given that this epidemiologic study did not modify existing diagnostic or therapeutic strategies.

Study Population

All consecutive patients meeting the following criteria were studied: the patient had been on mechanical ventilation (MV) for more than 48 h; had a new and persistent (> 24 h) infiltrate on chest X-ray; had macroscopically purulent tracheal aspirates; had a clinical status permitting flexible fiberoptic bronchoscopy; and VAP had been documented by microbiologic processing of protected specimen brush (PSB) or bronchoalveolar lavage (BAL) samples.

Diagnosis of VAP

Every patient clinically suspected of having pneumonia underwent fiberoptic bronchoscopy to obtain PSB and BAL samples before introduction of any new antibiotics. Specimens were collected and processed according to procedures extensively described previously (8). VAP was diagnosed based on PSB and BAL results according to the following "significant" thresholds (two of the three following criteria): PSB cultures yielding  10³ colony-forming units (cfu)/ml;  2% of the recovered cells containing intracellular bacteria (ICB) upon direct examination of the BAL fluid (BALF); or BAL cultures yielding  10⁴ cfu/ml. Bacterial identification and susceptibility tests were performed using standard methods. For the purpose of the study, only bacteria that grew at significant concentrations were considered to be causative microorganisms.

Other Microbiologic Specimens

For each VAP episode, all microbiologic specimen results available in the patient's chart on the day of fiberoptic bronchoscopy were prospectively reviewed and recorded. During the study period, all patients admitted to the ICU were screened with nasal and rectal swabs within 24 h of admission and weekly thereafter to identify S. aureus and A. baumannii carriages. Urine samples were systematically cultured at admission, every Monday, and when urinary tract infection was clinically suspected. Central venous catheters, arterial catheters, and Swan and Ganz catheters were systematically cultured at removal. For postoperative cardiac surgery patients with acute bacterial mediastinitis, mediastinal drainage fluids were cultured three times a week until removal of drainage tubes. All other microbiologic specimens, for example, blood cultures, and other miscellaneous specimen cultures were obtained based on clinical suspicion of infection. No bronchopulmonary samples (BAL, PSB, and tracheal aspirates) were obtained at predetermined times, but only when they were considered clinically justified by the team of physicians in charge of the patients. However, our general policy is to maintain a very high index of clinical suspicion in all patients who are mechanically ventilated in our ICU, in order not to miss any episode of VAP. This is why a large number of pulmonary specimens were obtained from ventilated patients during the study period. Microorganism susceptibilities were determined using the criteria established by the "Comité National de l'Antibiothérapie," the official French committee responsible for this classification.

The species and antimicrobial susceptibilities of isolated microorganisms responsible for VAP were compared with those previously recovered. Bacteria of the same genus and species and the same antibiotic susceptibility patterns were considered "identical."

Data Collection

For each episode of documented pneumonia, the following clinical variables were recorded on the day of bronchoscopy: temperature; white blood cell count (WBC)/ml; ratio of arterial oxygen tension to fraction of inspired oxygen (Pa_O2/FI_O2); radiologic score according to the definition previously used by Fagon and coworkers (9); duration of MV before nosocomial pneumonia; number of causative organisms; and the presence or absence of any antimicrobial agents for more than 24 h during the 15 d preceding the episode.

Costs associated with routine microbiologic surveillance were estimated by multiplying the charges associated with each microbiologic procedure, as determined by the French Social Security system, by the numbers of procedures performed (10).

Statistical Analyses

Data are reported as means ± SD. Sensitivity, specificity, positive and negative predictive values were determined for different groups according to standard definitions (11). The chi-square test or Fisher exact test was used to compare percentages. A value of p < 0.05 was considered to be statistically significant.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Population

Three hundred eighty episodes of suspected bacterial pneumonia were evaluated using fiberoptic bronchoscopy during the study period. However, on the basis of the criteria previously defined, VAP was microbiologically confirmed for only 91 patients, for a total of 125 episodes (20 patients had two episodes, 4 had three episodes, and 2 had four episodes). Demographic characteristics and other markers of severe illness at ICU admission for the 91 patients prospectively studied for the development of VAP are reported in Table 1. Forty-four patients survived and 47 died, for a case-fatality rate of 52%. The clinical characteristics at the onset of the 125 VAP episodes are summarized in Table 2. Antibiotics had been administered during the 15 d preceding VAP onset for 109 (87%) of the 125 episodes (Table 2).

A total of 220 microorganisms were grown at significant concentrations from PSB or BALF samples for the 125 VAP episodes and considered responsible for pneumonia (Table 3). Sixty-one (49%) of the 125 VAP episodes were polymicrobial.

Prior Microbiologic Specimens and Culture Results

A total of 5,576 microbiologic specimens (mean 44.6 ± 37.9 per episode) had been obtained before VAP onset, and their results were available on the day of bronchoscopy (Table 4). Among them, 914 (16%) grew at least one microorganism, for a total of 445 bacteria. Systematic surveillance cultures (n = 732) represented 13% of all recorded samples, but yielded only 58 (8%) positive cultures. Previous respiratory secretion cultures were available for 102 (82%) of the 125 VAP episodes, for a total of 366 specimens (mean, 3 ± 3 per episode), including 37 tracheal aspirates, 182 BAL specimens, and 147 PSB samples. Of these specimens, 63% grew at least one microorganism. The total costs associated with these 5,576 specimens were estimated to be 173,482 euros (Table 4).

Comparison of Microbiologic Findings

Among the 445 microorganisms isolated from the various microbiologic specimens obtained before VAP onset, 73 (16%) were also retrieved from bronchoscopic specimens and considered to be responsible for pneumonia, whereas 372 (84%) were not. Therefore, of the 220 microorganisms responsible for VAP, 147 (67%) had not been previously recovered. The 73 VAP-causative microorganisms were isolated from different bacteriologic specimens and sometimes from multiple sites in the same patient: 53 from respiratory secretion cultures, 21 from catheter-tip cultures, 17 from routine surveillance cultures, 8 from urine cultures, 7 from blood cultures, and 15 from miscellaneous cultures (operative sites, 12; and other cultures, 3). No specific site had a predictive value for predicting subsequent pneumonia caused by the same microorganism that exceeded 18%, ranging from 5% for microorganisms isolated from blood cultures to 18% for those isolated from bronchopulmonary samples. Interestingly, however, when one microorganism was isolated from multiple sites including a bronchopulmonary specimen, its predictive value was higher than when it was isolated from multiple sites but not from the lung, or from only a single nonpulmonary site (31%, 7%, and 5%, respectively, p < 0.01).

Because we were using bronchoscopic specimens to identify VAP-causative microorganisms, we may have either considered as pathogens some contaminants originating from the upper airways or, on the other hand, missed a certain proportion of pathogens. Data were reanalyzed two times: first, after exclusion of low-pathogenicity microorganisms, such as streptococci, Neisseria, and coagulase-negative staphylococci; and second, after inclusion of bacteria that grew at nonsignificant concentrations in bronchoscopic specimens. In the first case, the total number of microorganisms considered as VAP-causative decreased from 220 to 165, whereas it increased to 266 in the second case. However, whatever the definition used to identify VAP-causative microorganisms, the proportion that had been previously isolated by routine microbiologic culture results did not change significantly (42% and 33%, respectively, versus 33% using the usual definition, p = not significant [NS]).

As indicated in Figure 1, specimen culture results obtained before the onset of VAP were noncontributive for the selection of initial antimicrobial therapy for a total of 47 (38%) VAP episodes: 17 because they were all sterile, and 30 in which all previously isolated bacteria were not involved in the subsequent lung infection and were thus misleading. All the microorganisms ultimately responsible for VAP were previously recovered for only 39 (31%) episodes. Interestingly, the total number of microbiologic specimens that had been obtained before VAP onset was statistically lower for the 47 VAP episodes for which no causative microorganism was isolated before, compared with the 78 episodes for which at least one causative microorganism was previously isolated (mean 23 ± 17 versus 58 ± 41 per episode, respectively, p < 0.001).

View larger version (36K): [in this window] [in a new window]   Figure 1.   Contribution of prior microbiologic specimens to the identification of causative microorganisms in 125 episodes of VAP. The total numbers of specimens obtained before VAP onset for each of the four categories of VAP episodes represented were 193, 874, 2,144, and 2,365, respectively.

The proportion of VAP-causative microorganisms that were isolated by routine microbiologic specimens in the 16 patients who did not receive any antibiotics before VAP onset was 16% (5 of 32), compared with 36% (68 of 188) in the 109 patients who had received prior antibiotics (p = 0.03). This difference in the potential value of routine microbiologic specimen culture results for predicting VAP-causative microorganisms in patients with or without prior antibiotics was explained by the lower duration of MV (7 ± 4 versus 24 ± 20 d, p = 0.0006) and the lower number of microbiologic specimens that had been obtained before VAP onset (mean 14 ± 6 versus 49 ± 39 per episode, p = 0.0004) in patients without prior antibiotics than in patients with prior antibiotics.

Similarly, although only 20 (18%) of the 112 VAP-causative microorganisms were previously isolated in the 60 episodes for which prior duration of MV was < 15 d, this proportion was 49% (53 of 108) in the 65 episodes for which prior duration of MV was  15 d (p < 0.0001). As for the subgroup of patients who had not received any antibiotics before VAP onset, this lower usefulness of routine microbiologic results in patients with the shortest duration of MV was explained by the lower number of microbiologic specimens that had been obtained before VAP onset, compared with patients for whom prior duration of MV was  15 d (mean, 22 ± 19 versus 66 ± 39 per episode, p < 0.0001).

Prior Respiratory Specimens

Prior respiratory secretion culture results (mean, 4 ± 3 per episode) were available for 102 (82%) VAP episodes (mean time before VAP onset, 8 ± 9 d). In this subgroup, all prior respiratory microbiologic samples were sterile for 12 (12%) episodes, or the isolated bacteria were misleading for 26 (26%). At least one, but not all, VAP-causative bacteria were isolated from prior respiratory specimens for 28 (27%) episodes, whereas all VAP-causative bacteria were previously isolated for only 36 (35%) episodes. To determine whether the time elapsed between the last respiratory specimen obtained and the day of bronchoscopy documenting VAP helped to identify the VAP-responsible organisms in the event of subsequent lung infection, we compared their contributive usefulness according to this variable (Figure 2). All the bacteria responsible for VAP were recovered before VAP onset for 17 (52%) of the 33 episodes for which the last respiratory specimen had been obtained less than 72 h earlier, whereas this was the case for only 19 (28%) of the 69 episodes for which the last microbiologic specimen preceded pneumonia by more than 72 h (p < 0.03).

View larger version (27K): [in this window] [in a new window]   Figure 2.   Contribution of prior microbiologic respiratory specimens to the identification of causative microorganisms in 102 episodes of VAP. Black bars correspond to episodes for which at least one prior respiratory specimen was obtained less than 72 h before VAP onset, cross-hatched bars to episodes for which prior respiratory specimens predated VAP by 4 to 7 d, and gray bars to episodes for which prior respiratory specimens predated VAP by more than 7 d.

Similarly, the positive predictive value of a microorganism isolated from respiratory secretions for predicting subsequent pneumonia caused by the same organism was better when it was isolated within 72 h of VAP than when this interval was greater than 72 h (56% versus 13%, respectively; p = 0.001). Of the 182 BAL specimens that were obtained before VAP onset, 141 (77%) grew at least one microorganism, compared with only 61 (41%) of the 147 PSB specimens. However, no differences were noted in the operative indices of microorganisms recovered by bronchoscopic specimens according to the technique used for recovering pulmonary secretions. For example, the positive predictive value of a microorganism recovered by BAL (25%) was very similar to that of one recovered by PSB (28%, p = NS).

Potentially Drug-resistant Microorganisms

Table 5 details the diagnostic usefulness of previous colonization or infection with MRSA, P. aeruginosa, or A. baumannii for predicting subsequent pneumonia caused by these microorganisms in terms of sensitivity, specificity, and predictive values. Prior microbiologic specimen positivity for one of these bacteria was relatively specific (84%, 85%, and 79%, respectively) for subsequent pneumonia caused by the same microorganism, but sensitivities never exceeded 70% (66%, 47%, and 35%, respectively). Therefore, in a clinical setting in which the frequencies of VAP caused by potentially drug-resistant MRSA, P. aeruginosa, and A. baumannii, were high (28%, 26%, and 16%, respectively), the positive predictive values for prior specimens from which one of these microorganisms had been recovered were 62%, 52%, and 24%, respectively, and their respective negative predictive values were 86%, 82%, and 87% (Table 5). This somewhat better correlation observed between P. aeruginosa, MRSA, and A. baumannii and VAP-causative microorganisms was not explained by the fact that they were disproportionately isolated within 72 h of VAP. In fact, these three potentially drug-resistant microorganisms represented 20 (24%) of the 83 microorganisms isolated within 72 h of VAP onset and 90 (22%) of the 405 microorganisms isolated more than 72 h before VAP onset (p = NS).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

To test the hypothesis that results of routine microbiologic cultures of specimens obtained before VAP onset in ICU patients may help to predict the causative microorganisms and thus to select effective initial antimicrobial therapy, we undertook a prospective study on patients who required MV for more than 48 h and for whom strict bronchoscopic criteria were applied to diagnose pneumonia and identify the causative pathogens. Although a large number of various prior microbiologic specimen culture results (mean, 45 ± 38) were obtained before bronchoscopy for each VAP episode, only 73 of the 220 VAP-causative microorganisms were isolated by these routine analyses and their susceptibility patterns available to guide initial antimicrobial treatment. Specimen culture results obtained before VAP onset were noncontributive for selection of initial antimicrobial therapy for a total of 47 (38%) episodes, and all the microorganisms ultimately responsible for lung infection were previously recovered for only 39 (31%) episodes, making prospective identification of the true pathogens and the choice of effective empiric antibiotic therapy possible only for the latter. Even for patients for whom an available prior respiratory specimen culture result predated VAP onset by less than 72 h, all causative pathogens were recovered for only 52% of them. Finally, the diagnostic usefulness of previous colonization or infection with MRSA, P. aeruginosa, or A. baumannii, as documented by routine microbiologic culture results, for predicting subsequent lung infection caused by these microorganisms, appears limited. Although prior microbiologic specimen culture results for one of these bacteria were relatively specific for subsequent pneumonia caused by the same microorganism, with less than 20% false-positive results, sensitivities never exceeded 70%, rendering difficult the use of only these data for the selection of initial VAP therapy in a clinical setting in which the frequencies of lung infection caused by these potentially drug-resistant microorganisms were high.

Routine surveillance for bacterial colonization has been used for the past three decades as a tool for the identification of nosocomial infection-causative pathogens; however, to the best of our knowledge, its usefulness has never been proven (12, 13). For example, Finelli and coworkers conducted a prospective cohort study on 154 low-birth-weight infants who were admitted to a neonatal ICU, to examine the utility of surveillance for detection of pharyngeal colonization and control of severe bacterial infection (12). They concluded that pharyngeal surveillance cultures provide little clinically meaningful information and have no apparent utility in the neonatal ICU setting.

Several factors may explain the lack of accuracy of routine microbiologic specimen culture results for predicting the causative microorganisms of pneumonia in ICU patients requiring MV. First, the role played by colonization of some of the sites sampled before VAP onset, such as the nares, skin, or urine, in the pathogenesis of nosocomial lung infection, is probably limited, thus explaining the absence of concordance between these microorganisms and those responsible for VAP. Second, a very large number of different bacterial species were recovered from specimens obtained before VAP onset, whereas only a much smaller number of microorganisms were responsible for the infection, making identification of the "true" VAP pathogens difficult. Finally, even when bacteria were isolated from a site susceptible of having a role in lung infection, such as the tracheobronchial tree, the interval between prior specimen culture results and VAP onset was frequently long enough to permit the development of lung infection caused by microorganisms other than the ones previously isolated.

When the analysis focused on VAP episodes for which prior respiratory secretion culture results were available, hypothesizing that this microbiologic information might be particularly useful for identifying the responsible organisms in the event of subsequent pneumonia, results were still disappointing. Even when a prior respiratory specimen culture result was available less than 72 h before VAP onset, all causative pathogens were recovered for less than 60% of these episodes. Although many investigators have clearly documented that tracheal colonization by potentially pathogenic microorganisms precedes lung infection in a majority of, but not all, ventilated patients (14), recent data have also emphasized that the pattern of tracheobronchial colonization, and especially the types of microorganisms involved, is a dynamic process, with rapid modification of the bacterial flora present at that level according to factors such as prior duration of MV and prior antibiotics (17, 18).

In a recent study, Delclaux and coworkers prospectively evaluated lower respiratory tract colonization and infection in 30 patients with severe acute respiratory distress syndrome (ARDS), using repeated quantitative cultures of plugged telescopic catheter (PTC) specimens taken blindly through the endotracheal tube every 48 to 72 h after ARDS onset (19). Lower respiratory tract colonization, defined as a PTC specimen growing < 10³ cfu/ml or  10³ cfu/ml in the absence of clinical criteria of VAP, was consistently followed by infection with the same microorganisms; however, colonization preceded BAL microbiologically confirmed VAP in only 67% of the VAP episodes. Therefore, careful evaluation of distal airways colonization can fail to document at least one-third of VAP episodes. Furthermore, it should be emphasized that these results were obtained using protected catheters to assess distal airways colonization and not tracheal aspirates; that technique protects against upper airways contamination of the sample, and organisms recovered better reflect those present in the lower airways that are more likely to cause infection.

Routinely harvesting and culturing tracheal aspirates can increase the number of microorganisms isolated before VAP onset that are not responsible for VAP and, thus, make the choice of the optimal antimicrobial therapy even more difficult. Such a strategy may also considerably increase the work of the microbiology department without having any positive impact on patient management. Accordingly, and even when colonization of the upper airways occurred in most patients who subsequently developed VAP, most data suggest that a strategy based on routine tracheal microbiologic surveillance for identifying VAP-causative microorganisms would be of limited value, particularly when tracheal aspirates, and not distal specimens, are used, and when respiratory specimens are not obtained repeatedly at very short intervals, i.e., every 48 to 72 h.

Several investigations have clearly demonstrated that colonization with potentially drug-resistant pathogens, such as MRSA or extended-spectrum beta-lactamase-producing strains of Klebsiella pneumoniae, is associated with an increased risk of infection caused by the corresponding microorganism (7, 20). These results were confirmed in our study; positive predictive values of recovering such a microorganism from a specimen were 62%, 52%, or 24% for VAP caused by MRSA, P. aeruginosa, or A. baumannii, respectively. However, because the sensitivity of prior microbiologic culture results for identifying bacteria causing VAP never exceeded 70% (66%, 47%, and 35%, for MRSA, P. aeruginosa, and A. baumannii, respectively), selection of initial antimicrobial therapy for patients with VAP can hardly be based only on these results, especially for deciding to use (or not) vancomycin or a broad-spectrum beta-lactam effective against P. aeruginosa or A. baumannii.

Certain limitations to the microbiologic aspects of this study, however, merit attention. First, no genotypic analyses of the different strains that were isolated were performed, which may have falsely increased the proportion of VAP-causative microorganisms considered as having been previously identified. Second, microbiologic features associated with VAP in this study reflect the type of patients admitted to our ICU and, therefore, may not be applicable to all patients receiving MV. Third, our ICU accumulates risk factors for an increased rate of infection caused by multiresistant pathogens (24). Finally, that repeated cultures of respiratory tract specimens were not performed at predetermined times may have restricted the potential usefulness of routine microbiologic culture results in our study.

In summary, the findings of this study indicate that serial routine microbiologic specimen culture results obtained before VAP onset can identify only a small percentage of the causative microorganisms. Furthermore, these specimens are often misleading for the selection of initial antimicrobial treatment, because they mostly isolate bacteria that are not involved in the subsequent lung infection. Most episodes of infection caused by MRSA, P. aeruginosa, or A. baumannii were preceded by prior isolation of these bacteria, but because the sensitivities of prior microbiologic culture results for identifying these microorganisms never exceeded 70%, selection of initial antimicrobial therapy for patients with VAP can hardly be based only on results of routine microbiologic specimens. However, when one of those three microorganisms (or any pathogen) is isolated from respiratory secretions within 72 h of VAP, it should be covered by the antimicrobial regimen selected, though predictive values do not certainly exceed 50% to 60%.