INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The choice of initial antibiotics in patients with nosocomial pneumonia before etiological diagnosis is available remains a challenge. In 1989, Celis and coworkers (1) reported that an inappropriate antibiotic regimen was associated with a fivefold increase in crude mortality. A recent study (2) has confirmed that in patients with ventilator-associated pneumonia (VAP) an inappropriate initial antibiotic choice is associated with a significant increase in both crude and related mortalities. It therefore appears that the survival of intubated patients with pneumonia could be increased by improving the process of selecting initial antimicrobial regimens. The American Thoracic Society (ATS), well aware of the critical importance of the topic, created a task force to give recommendations for empirical treatment of nosocomial pneumonia (including pneumonia in intubated patients) based on the opinions of recognized experts (3). A further report from Europe (4) described the distribution of causative organisms according to easily identifiable risk factors and could complement the ATS guidelines. These investigators suggested that classifying patients according to prior duration of mechanical ventilation MV (< 7 or  7 d) and prior use or nonuse of antibiotic in the past 15 d provided a rational basis for selecting the initial therapy of patients suspected of having VAP. Whether this type of evidence is useful in other institutions remains unclear.

The present study is a retrospective analysis of information recorded prospectively for other purposes regarding all patients admitted to three intensive care units (ICUs) who developed VAP. This information enabled us to assess the usefulness of the epidemiologic characteristics described by Trouillet and coworkers (4) in predicting etiologies of VAP in other hospitals. In contrast to episodes of community-acquired pneumonia, the incidence of multiresistant pathogens varies widely, and is closely linked to local factors. We therefore hypothesized that empirical initial antibiotic choices for VAP should be based on up-to-date information of local epidemiology instead of general recommendations based on other institutions.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Population

The study was conducted in three ICUs: at the Hospital de Clinicas, in Montevideo, Uruguay; at the Hospital Virgen del Rocio, in Seville, Spain; and at the Hospital de Sabadell, in Barcelona, Spain. All patients admitted to these three ICUs requiring MV for more than 48 h in the period 1994 through 1996 were included in the study and prospectively followed. In Barcelona some patients received subglottic secretion drainage (5), whereas in Seville all subjects received a short antibiotic course of antibiotics for perioperative infection prophylaxis. During the study period, all patients who developed pneumonia in these ICUs were initially eligible and evaluated by three of the investigators (R.L., H.C., J.R.), as part of a prospective VAP study. A partial overlap existed between some patients in Barcelona or Montevideo and two previous reports (2, 6) with completely different aims. All received sucralfate, antiacid, and/or H₂ blockers, and no selective decontamination regimen was applied.

A diagnosis of pneumonia was considered when new, persistent pulmonary infiltrates not otherwise explained appeared on chest X-rays. Moreover, at least two of the following criteria were also required: (1) fever  38° C; (2) leukocytosis  10,000 per mm³; (3) purulent respiratory secretions. A pneumonia was considered ventilator-associated when it occurred after intubation and was judged not to have been incubated before starting MV (7). Fiberoptic bronchoscopic examination using a protected specimen brush (PSB) or bronchoalveolar lavage (BAL) was performed on each of these patients within the first 24 h of the development of a new pulmonary infiltrate. The diagnosis of VAP was upheld only if the PSB or BAL yielded  1,000 colony-forming units (cfu)/ml or 10,000 cfu/ml, respectively, of at least one microorganism. Episodes with microorganisms under these thresholds were excluded because of low specificity. Episodes in which no etiologic diagnosis was established were also excluded. Bacterial identification tests were performed using standard methods. VAP was considered to be caused by "potentially resistant" bacteria when methicillin-resistant Staphylococcus aureus (MRSA), Pseudomonas aeruginosa, Acinetobacter baumannii, or Stenotrophomonas maltophilia grew at significant concentrations from PSB or BAL specimens. For the purpose of this study, only the first episode of nosocomial pneumonia was taken into account.

To analyze the predisposing factors for developing VAP, the following variables were recorded: age, sex, underlying disease, severity, prior trauma or surgery, the presence of chronic obstructive pulmonary disease (COPD), other pulmonary diseases, cause of intubation, septic shock, adult respiratory distress syndrome (ARDS), the duration of MV prior to the development of VAP, and prior pneumonia. Previous antibiotic therapy was only considered when a patient received antimicrobial agents (at least 24 h) during the 15 d before the pneumonia episode. In all these cases, the class of antibiotic was recorded. Patients receiving imipenem, third-generation cephalosporins, or fluoroquinolone were grouped in the category of "broad-spectrum antibiotics." COPD was diagnosed using the standard criteria recommended by the American Thoracic Society (8). Severity of underlying disease conditions was evaluated with Acute Physiology and Chronic Health Evaluation II (APACHE II) score (9) for each patient in the first 24 h after ICU admission. Surgery or multiple trauma were considered to be present if they took place within 2 wk before the episode. Multiple trauma was defined as presence of injury to more than one body area or system. The joint category grouped the isolates from the three current treatment sites.

Microbiology

Fiberoptic bronchoscopic examination was performed using the PSB technique or BAL to obtain uncontaminated lower airway secretions for bacterial cultures, as previously reported (10). Specimens were transported to the laboratory immediately after collection. The vial was then vortexed vigorously for at least 60 s to thoroughly suspend all material from the brush. Two serial 100-fold dilutions were made, and 0.1-ml aliquots of the original suspension and each dilution were inoculated on appropriate plates. Two serial 10-fold dilutions were then done on the recovered BAL fluid, and 0.01-ml aliquots of the original suspension and each dilution were placed onto plates in the same way as for the PSB sample. In accordance with the standards adopted in previous studies, bacterial counts  1,000 cfu/ml (PSB) or 10,000 cfu/ml (BAL) were taken as the cutoff points for the pulmonary infection diagnosis (11). Bacterial identification and susceptibility testing were performed by standard methods (12).

Definition of Groups of Patients

Four groups of patients were defined in accordance with the criteria suggested by Trouillet and coworkers (4). In brief, Group 1 included patients ventilated for < 7 d without any prior antibiotic therapy. This group is similar to patients also classified in Group 1, according to the ATS guidelines (3), who are likely to be infected by the core organisms. Patients included in the current Groups 2, 3, and 4 are classified in Group 3 of the ATS guidelines (3) being at risk for potentially resistant organisms.

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 days. Group 4 included patients ventilated for  7 d who had received antibiotic therapy within the previous 15 d.

Statistical Analysis

Contingency tables were analyzed using the two-tailed chi-square test. When not appropriate, Fisher exact test was used. Continuous variables were compared using Student's t test; when not appropriate, the Mann-Whitney U-test was used. Differences between groups were considered to be significant for variables yielding a p value  0.05.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

A total of 321 significant concentrations of bacteria were cultured in 290 episodes of VAP. The distribution of the causative bacteria among the institutions from Barcelona (B), Montevideo (M), and Seville (S) was 143, 123, and 55, respectively. The most frequently isolated organisms were P. aeruginosa (31.7%), methicillin-sensitive Staphylococcus aureus (MSSA) (11.8%), A. baumannii (11.8%), Haemophilus influenzae (8.4%), Streptococcus pneumoniae (7.7%), and MRSA (3.1%). No infections caused by MRSA were documented in our cohort of patients who had not recently received antimicrobial treatment (Group 1 or 3). In addition, Enterobacteriaceae as a group represented 13.7% of isolates.

Epidemiologic characteristics at the time of ICU admission of the patients in the different institutions, including demographic information, indication for MV, presence of comorbidities, and severity of illness at admission are detailed in Tables 1 and 2. Most of these characteristics were significantly different when compared with the reference site (Paris). All patients in Seville underwent cardiovascular surgery, whereas the percentage of postoperative patients in Barcelona or Montevideo was lower than 40%. Similarly, no patients had prior pneumonia, sepsis, or septic shock in Seville, whereas these conditions were common in Barcelona, particularly in patients who developed VAP by "potentially resistant" bacteria (Table 1). Half of patients presented trauma in Montevideo, whereas this condition was unlikely in Barcelona and not present in Seville. Finally, mean APACHE II was significantly higher in Paris when compared with the other three sites.

Group 1 included 89 pathogens isolated from 85 patients ventilated for < 7 d without any prior antibiotic therapy (excepting short-term surgical prophylaxis). The distribution of causative bacteria among the institutions and the reference site (4) is detailed in Table 3. A statistically significant (p < 0.05) difference between the incidence of multiresistant pathogens in the joint group and the reference group was observed. This difference was caused by the isolation of some nonfermentative gram-negative bacilli in the three current treatment sites, mainly A. baumannii (p < 0.05) in Montevideo. Indeed, P. aeruginosa remained extremely infrequent (3 of 89, 3.3%) in Group 1. In addition, Barcelona showed a significantly (p < 0.05) lower incidence of Enterobacteriaceae, although no significant differences were documented when comparing the joint group with the reference site regarding these pathogens. In contrast, Neisseria sp only was isolated in Paris.

Group 2 also included 89 pathogens isolated from 81 patients ventilated for < 7 d who had received at least one antibiotic within the previous 15 d. Thirty-eight (46.9%) of these patients had received broad-spectrum antibiotics (imipenem, third-generation cephalosporin, or fluoroquinolone). Table 4 shows the distribution of organisms between these treatment sites in Group 2. Nineteen isolates of P. aeruginosa were found in Barcelona, compared with only four in Paris (p < 0.05). Although no episodes resulting from MSSA or S. pneumoniae were documented in Barcelona or Paris, both pathogens were isolated in Montevideo and Seville; indeed, the incidence of these organisms was significantly different (p < 0.05) when the joint category was compared with Paris. Methicillin-sensitive S. aureus was particularly frequent in Seville and this pathogen was significantly (p < 0.05) more frequent than in Paris when compared individually.

Group 3 included 25 pathogens isolated from 22 patients ventilated for  7 d who had not received antibiotics during the 15 preceding days. Table 5 detailed the distribution of isolates in the different sites for Group 3. The presence of P. aeruginosa remained significantly (p < 0.05) higher in Barcelona than in Paris, and this finding meant that the joint category presented statistically significant differences with respect to the reference site. On the other hand, Paris showed a significantly higher (p < 0.05) incidence of MSSA, "other streptococci" and Neisseria sp. Conversely with this primary endogenous flora, the incidence of H. influenzae was significantly lower (p < 0.05) than in the joint category.

Finally, Group 4 included 118 pathogens isolated from 110 patients ventilated for  7 d who had received antibiotic therapy within the previous 15 d. Forty-nine (44.5%) of these patients had received at least one broad-spectrum antibiotic. Table 6 shows the distribution of organisms by study site in Group 4. In Barcelona, P. aeruginosa remained significantly (p < 0.05) more frequent than in Paris, whereas the other multiresistant pathogens were extremely infrequent (p < 0.05). Half the episodes in Seville were due to A. baumannii, and the difference in distribution vis-à-vis Paris was therefore significant (p < 0.05). Likewise, most episodes (90.9%) of infection caused by A. baumannii in the reference site were found in Group 4, but this infection represented only 42% of isolates in the joint group (p < 0.05). Whereas some episodes due to H. influenzae were reported in Paris, this pathogen was never isolated in group 4 at the other sites (p < 0.05). Similarly, "other streptococci" remained uncommon (3/118 versus 14/152, p < 0.05).

When compared with the ATS guidelines for empiric treatment of nosocomial pneumonia, patients included in Groups 2, 3, and 4 in our current report will be enclosed in the Group 3 suggested by the ATS guidelines (3). This is summarized in Table 7. When only episodes occurring within < 5 d of intubation were considered, as suggested for Group 1 of the ATS guidelines, only seven multiresistant organisms were isolated (Table 8). Three episodes caused by P. aeruginosa were documented in patients with COPD and the remaining four episodes were caused by A. baumannii.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This study demonstrates that initial antimicrobial therapy for patients with VAP should vary markedly according to site. When the distribution of the causative bacteria isolated from the three current sites was analyzed according to Trouillet and coworkers' four groups of patients (4) (a classification made on the basis of prior duration of MV and presence or absence of antibiotics preceding the event), substantial differences were documented in all groups of organisms. No clinically significant site differences existed for Group 4, and the incidence of VAP caused by potentially drug-resistant bacteria remained high (77.9%). However, for patients with early-onset pneumonia who had received prior antibiotics (Group 2) and patients with late-onset pneumonia who had not received prior antibiotics (Group 3), empirical combination therapy as suggested by Trouillet and coworkers (4) may be inappropriate owing to the possible presence of MRSA or A. baumannii in certain sites. Moreover, this pathogen may even be present in Group 1. These variations in organisms across treatment sites can be explained by differences in patient demographic characteristics or comorbidities (Tables 1 and 2), strategies for pneumonia prophylaxis, and particularly local patterns of resistant organisms. Thus, subsequent decisions regarding initial antibiotic choices should consider their own local patterns. Prescriptions ignoring the variability of pathogens linked to exogenous acquisition and based on other considerations run an enormous risk of failure or unnecessary expense.

Prior studies (13) have suggested that the main variables determining the causative pathogen in VAP are: underlying disease (and comorbid conditions), exposure to risk (intubation period), and selection of flora by systemic drugs (particularly antimicrobial agents). This is true for pathogens of endogenous origin (e.g., MSSA, H. influenzae, Enterobacteriaceae and P. aeruginosa), as has been extensively reported (14). In addition, factors associated with transmission to patients and environmental contamination in a given institution are implicated in colonization/infection by organisms of exogenous origin. Thus, variations in local patterns of environmental contamination will lead to large differences in prescribing practices.

The outstanding finding of this study is that the distribution of VAP due to A. baumannii varies markedly among the four treatment sites. Whereas up to 90% of episodes in Paris were confined to Group 4, in the joint category more than 50% were documented in the other groups. Recent reports have found that A. baumannii may rapidly colonize patients who were admitted to ICUs when infection is endemic (16) or may even have been acquired in other hospital areas before ICU admission (17). Other investigations reported that no association with prior antibiotic use was needed (5, 16). However, this study suggests that prior antibiotic exposure is required for emergence of S. maltophilia. Our current findings suggest that the decision to cover A. baumannii when VAP is suspected should be customized to each institution.

P. aeruginosa, the remaining nonfermentative bacilli, was the most frequently documented pathogen in the joint category; its epidemiologic pattern differs from those of A. baumannii and S. maltophilia. This was confirmed in a recently published study (5) which stated that the risk factors for VAP caused by A. baumannii differed substantially from those associated with VAP caused by P. aeruginosa (15). Indeed, the current study revealed that P. aeruginosa remained extremely infrequent (3.3%) in Group 1, supporting the suggestion that antipseudomonal coverage is not required for patients who develop pneumonia within the first week of ventilation and have not received prior antibiotics (4, 15).

This study confirms that MRSA is extremely unlikely to be found in absence of exposure to a prolonged antimicrobial regimen. Only one of the 42 isolates (in Paris) was reported in this setting. Under selective antibiotic pressure, colonizing flora change rapidly and subsequent courses of broad-spectrum antibiotics further select and amplify the colonizing MRSA population (18). In a study of S. aureus VAP (19), all patients with MRSA VAP had recently received antibiotics, compared with only 21% of those with MSSA episodes. For ventilated patients who develop pneumonia, all these findings suggest that vancomycin should not be prescribed if patients have not previously been exposed to antimicrobial therapy.

The study by Trouillet and coworkers (4) is a very important contribution to the literature because it is the first series that seeks to provide a more rational basis for selecting the initial therapy of patients in whom VAP is suspected based on evidences. What we have learned from comparing Trouillet's results with those found at the three new treatment sites is that the epidemiologic variables that determine the presence of P. aeruginosa or endogenous flora did not significantly differ in the other sites. This observation contrasts with the evidence that the resistant population of organisms acquired by an exogenous route has an epidemiologic pattern that differs from institution to institution. Prescribing practices for VAP should therefore take these variations into account.

The ATS recommendations (3) only defined two ICU groups. Early onset, no risk, and only at risk for the core organisms (Table 8) comprise Group 1 of the ATS recommendations. Our current data show that when only episodes occurring within < 5 d of intubation were considered, only seven multiresistant organisms were isolated. Three episodes caused by P. aeruginosa were documented in patients with COPD and this finding confirms previous observations (15). The remaining four episodes were caused by A. baumannii, suggesting that each hospital has its own unique bacteriology regarding organisms acquired by cross-contamination. The other three groups (Tables 4, 5, and 6) were all at risk for resistant pathogens: late onset with risks, or risks with any time of onset. This (Table 7) is Group 3 of the ATS recomendations (3). Once again, distribution of A. baumannii in the different sites differed and was independent of the stratification suggested by Trouillet and coworkers (4). On the other hand, this approach was extremely useful in identifying the subgroups of patients (Tables 3 and 5) in which MRSA are unlikely. Consequently, in absence of recent antibiotic use, even in late-onset episodes with risk factors, adding antibiotic coverage with vancomycin in the initial choice should not be considered.

In the current study, the demographics of the patient in each unit presented significant differences, as detailed in Tables 1 and 2. For example, all patients in Seville underwent cardiac surgery. In contrast, postoperative patients represented two-thirds of the study population in Paris and one-third in Barcelona and Montevideo. Half of patients presented trauma in Montevideo, whereas this condition was unlikely in Barcelona and not present in Seville. Probably these differences in host factors and the local epidemiology, are the most rational explanations for the discrepances observed.

General agreement exists that in research studies, the use of quantitative bronchoscopic techniques is of preference (20). Excluding patients with suspected pneumonia and low bacterial burden would perhaps limit the generalization of these results to ICUs that use quantitative culture techniques to diagnose and manage VAP. However, including episodes with lower bacterial burden or based on qualitative specimens would raise concerns regarding the true value of isolated organisms (11, 13). Avoidance of misclassification is essential to clarify the epidemiology and microbiology of VAP in ICU patients (21). Therefore, our approach provides the highest specificity to the isolated pathogens.

Finally, the current study has several limitations that should be borne in mind when interpreting the results. First, the sample is relatively small, and a type II error may be present in some of the comparisons. This is particularly true in Seville, and means that differences in etiologies between different sites may be even undervalued. Second, as Trouillet and coworkers (4) pointed out as a limitation in their study, length of stay in the hospital prior to the beginning of MV was not considered, and this variation may change episodes initially classified in Group 1 or 2 to Group 3 or 4. Third, the potential relationship between etiologies and different classes of antibiotics would not be investigated and this would be an additional factor selecting for specific multiresistant pathogens. Finally, all four treatment sites analyzed were located in teaching institutions and our observations may not be generalizable to all patients with VAP. Important intrahospital differences may be observed in large hospitals. Thus, in presence of several ICUs within one hospital, guidelines may require customization to each unit.

In summary, in deciding how to approach antimicrobial therapy for VAP, national or regional guidelines for initial antimicrobial therapy need to be modified to take into account local patterns of antimicrobial resistance. Likewise, health care systems will need to consider their own policies according to patient populations and local patterns of pathogen distribution in interpreting the way that national guidelines are implemented in their own institutions. Our comparative study suggests that regulations designed to deal with treatment based on other institutions are not likely to be either successful or cost-effective.