INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Nosocomial pneumonia is thought to be a common complication of the acute respiratory distress syndrome (ARDS). Most clinical studies have found that pulmonary infection affects between 34% and > 70% of patients with ARDS, often leading to the development of sepsis, multiple organ failure, and death (1). When the lungs of patients who have died of ARDS are examined histologically at autopsy, pneumonia may be demonstrated in as many as 73% (2).

The diagnosis of pulmonary infection in patients with ARDS, however, is often difficult. Several studies have clearly demonstrated the inability of physicians to accurately diagnose nosocomial pneumonia in this setting with the use of clinical criteria alone (1, 2, 7). Although it is difficult to distinguish clinically between bacterial colonization of the tracheobronchial tree and true pulmonary parenchymal infection, nearly all previous studies have relied solely on clinical diagnostic criteria, and therefore have probably included patients who did not have pneumonia, or have excluded patients with true pneumonia. Furthermore, most of these studies used the bacteriologic results obtained with tracheal secretions as their major source of culture material, despite the fact that the upper respiratory tracts of most ARDS patients are colonized with multiple pathogens. Alternatively, studies that restricted the diagnosis of pneumonia to nonsurvivors probably included only a limited portion of the clinical spectrum of the disease. Because of these potential limitations, previous epidemiologic investigations have not clearly defined the precise role of nosocomial pneumonia in the clinical course of ARDS.

Quantitative bacterial cultures of lower-respiratory-tract secretions collected during bronchoscopy with a protected-specimen brush (PSB) and/or by bronchoalveolar lavage (BAL) may enhance the accuracy of diagnosis of nosocomial pneumonia in mechanically ventilated patients (11, 12). Using such techniques at predetermined times from Days 3 to 21 after the onset of the syndrome in a large series of 105 patients with ARDS, Sutherland and colleagues recently concluded that the incidence of ventilator-associated pneumonia (VAP) may indeed be far lower than expected in this group of patients (9). Only 16 (15.2%) of their 105 patients met the quantitative criteria for pneumonia (PSB > 10³ cfu/ml or BAL  10⁴ cfu/ml), and no correlations were found between total colony counts in bronchoalveolar lavage fluid (BALF) or PSB cultures and severity of ARDS, as judged by Pa_O2/FI_O2 ratios, days on mechanical ventilation (MV), static lung compliance, or survival. Unfortunately, these results are probably not of general value, because most patients included in the study were lavaged while receiving antibiotics, and at predetermined times during the course of ARDS, rather than at the time of clinically suspected infection. For PSB and BAL sampling to be of diagnostic value, it is absolutely imperative that patients' antimicrobial regimens, if any, not be changed before pulmonary secretion specimens are obtained (11, 12). Even a few doses of an effective antimicrobial agent can rapidly decrease or even transiently render bacterial counts in PSB or BALF specimens negative, and thereby lead to a high number of false-negative results (13).

Therefore, the present study was undertaken to: (1) provide a comprehensive description of the epidemiologic and microbial etiologic aspects of microbiologically provable VAP in patients with ARDS, as determined through quantitative cultures of PSB and BALF specimens when potential confounding factors, such as those listed earlier, can be excluded; and (2) compare these characteristics with those present in patients without ARDS who also developed VAP. For this latter purpose we collected, via fiberoptic bronchoscopy and before any modification of existing antimicrobial therapy, PSB and BALF samples from every patient ventilated in our intensive care unit (ICU) who was clinically suspected of having developed nosocomial pneumonia.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The study was conducted between April 1, 1994 and June 30, 1995, in our ICU. Its protocol was in accord with the ethical standards of the Committee for the Protection of Human Subjects of the Hopital Bichat. In accordance with French law, no informed signed consent was mandatory, given that this epidemiologic study did not modify existing diagnostic or therapeutic strategy.

Study Population

All consecutive patients admitted to our ICU who received MV for  48 h during the 15-mo study period were included in the study. Patients with ARDS were prospectively identified when all the following criteria were met: (1) generalized pulmonary infiltrates involving at least three of four lung quadrants on chest radiographs; (2) a lung injury score (LIS) of greater than 2.5 based on blood-gas analysis, extent of edema on chest radiograph, positive-end expiratory pressure, and static lung compliance values (14); (3) an etiology compatible with ARDS; and (4) a pulmonary capillary wedge pressure (PCWP) < 18 mm Hg in those with pulmonary artery catheters, or no clinical or echocardiographic evidence of congestive heart failure (CHF) in those without.

No changes in the methods used to identify and manage nosocomial pneumonia were made during the study period (15). Because it may be difficult to exclude expansion of lung infiltrates in patients with diffuse bilateral infiltrates, these patients were categorized separately, and were kept under extreme vigilance for possible VAP throughout the study period. Fiberoptic bronchoscopy was therefore performed as soon as any such patient became febrile or deteriorated clinically, even when no modification of chest-radiographic findings could be ascertained. Other patients, with only limited prior radiologic abnormalities, were suspected of having developed nosocomial pneumonia when they developed a new pulmonary infiltrate or when an existing infiltrate progressed and had purulent tracheal secretions.

Therefore, all patients clinically suspected of having nosocomial pneumonia underwent immediate fiberoptic bronchoscopy, at which time specimens were obtained from the involved lung segment using a PSB and by BAL, before existing antimicrobial treatment (if any) was modified in any way. Only patients with a Pa_O2 < 70 mm Hg and an FI_O2 of 1.0, or with a very unstable hemodynamic status (hypotension with a systolic blood pressure < 90 mm Hg) despite optimal cardiovascular support, were excluded from the study. However, only BAL, and not PSB, was performed in patients with severe thrombocytopenia (< 50,000/mm³).

Specimen Collection

Before the administration of any new antibiotics, the following protocol was initiated in all patients included in the study. Patients were premedicated with intravenous midazolam and/or phenoperidine and paralyzed with a short-acting neuromuscular blocking agent (vecuronium bromide; Organon Teknika, Fresnes, France) 5 min before the sampling procedure; topical anesthesia was not used. The ventilator settings were adapted for better ventilation, and the FI_O2 was adjusted to 95% or more. The fiberoptic bronchoscope was introduced and advanced to the selected bronchial orifice on the basis of the radiographic location of a new pulmonary infiltrate or, in patients with diffuse bilateral infiltrates, into the area in which purulent secretions were most abundant. When diffuse endoscopic abnormalities were present in patients without clear-cut new pulmonary infiltrates, specimens were obtained from the right lower lobe. A PSB sample was retrieved using the previously described standardized technique (12). After the brushing, BALF was performed by instilling a total of six 20-ml aliquots of sterile, nonbacteriostatic saline. The liquid recovered after the first aliquot was discarded, and the remaining lavage fluids were pooled and passed through sterile gauze. The entire procedure was well tolerated by all the patients, with the only significant complications being a transient oxygen desaturation in two patients with ARDS and benign bronchial bleeding in one patient without ARDS.

Microbiologic Processing

BAL fluid and PSB samples. The BALF was divided into two samples: one for direct examination and the other for quantitative culture. Total cell counts were made, after centrifugation, on an aliquot of the original resuspended, pooled fluid. Cytocentrifuge preparations were made with a Cytospin 2 centrifuge (Shandon Southern Products, Cheshire, UK) and stained with a modified Wright-Giemsa stain (Diff-Quik; Baxter Diagnostic AG, Düdingen, Switzerland). Differential cell counts and the percentages of cells containing intracellular bacteria (ICB) were determined as previously described (11). The specimens for culture were processed within 60 min after the BAL procedure, and remained at room temperature during that time. After the tubes had been vortexed for 60 s, 1 ml of undiluted BALF was used to prepare serial dilutions, and each dilution was plated for aerobic and anaerobic cultures on blood chocolate, blood-charcoal-yeast extract with and without antibiotics, and Sabouraud media. The tip of the PSB was cut, dropped into 1 ml of sterile water, and vortexed for 1 min. Serial dilutions (to 10⁶) were then prepared, and each dilution was plated on the same media as for BALF cultures. Culture dishes for BAL and PSB samples were evaluated for growth after 24 h and 48 h at 37° C. The numbers of bacteria in the original specimens were estimated by colony counts, and are expressed as cfu/ml of sample. Bacteria were identified with standard methods (16).

Patients who met at least one of the following three criteria were diagnosed as having microbiologically provable nosocomial pneumonia: > 5% of the cells in cytocentrifuge preparations of BALF contained ICB, as previously described in detail (11); at least one bacterial species grew at a significant concentration (> 10³ cfu/ml) from specimens obtained with a PSB, at least one bacterial species grew at a significant concentration (> 10⁴ cfu/ml) from BALF specimens, as previously described by ourselves and others (11, 12, 17). For the purpose of the study, only bacteria that grew at significant concentrations were considered to be causative microorganisms.

Data Collection

Upon the entry of each patient into the study, the patient's hospital chart was reviewed prospectively and the following clinical variables were recorded: age; sex; location before admission to the ICU; severity of the patient's underlying medical condition, stratified according to the criteria of McCabe and Jackson as rapidly fatal, ultimately fatal, or not fatal (18); primary admission diagnostic category; simplified acute physiologic score (SAPS II) (19); Glasgow coma scale (GCS) score; (GCS) presence or absence of infection and presence or absence of cardiovascular, respiratory, renal, hepatic, hematologic, and/ or neurologic dysfunction(s) diagnosed according to the definitions we previously established in the organ dysfunctions and infection (ODIN) model (20); and presence or absence of sepsis, severe sepsis, or septic shock (21). All of these parameters were recorded within the 24 h following admission of the patient to the ICU, with the worst value for each variable being retained.

Furthermore, for each episode of confirmed pneumonia, the following clinical variables were recorded on the day of bronchoscopy: duration of prior MV; temperature; blood leukocytes/mm³; Pa_O2/FI_O2; radiologic score (13); and presence or absence of any antimicrobial agents for > 24 h during the 15 d preceding the episode. Additionally, administration of each of the five following antibiotic classes during the 15 d preceding the event was recorded for each patient: imipenem, third-generation cephalosporin, aminoglycoside, fluroquinolone, and/ or "other" antibiotics (defined by exclusion). Broad-spectrum drugs (imipenem, third-generation cephalosporin, and fluoroquinolone) were regrouped to constitute a new variable. The number of antibiotic classes administered during the 15 d prior to fiberoptic bronchoscopy was also recorded.

Statistical Analysis

Clinical and laboratory data were analyzed statistically with Student's t-test for the comparison of continuous variables, the chi-square test or Fisher's exact text for the comparison of proportions, and analysis of variance (ANOVA) for the comparison of intergroup differences. Wilcoxon's rank test was also applied, to evaluate differences between groups when the distributions of the variables were not assumed to be normal. Time-to-event variables were estimated according to the Kaplan-Meier method, and were compared by means of the log-rank test. Statistical significance was defined as p < 0.05.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Population

Demographic characteristics and other markers of acute illness in the 243 patients studied for the development of VAP are shown in Table 1. Of these patients, 56 developed ARDS as defined by an LIS > 2.5 within the first 48 h of MV. No patients included in the control group subsequently fulfilled the criteria for ARDS. The syndrome was caused by a direct lung injury in 33 patients (community- or nosocomially acquired pneumonia, n = 29; aspiration of gastric contents, n = 3; near-drowning, n = 1), and was secondary to an indirect lung injury in 23 patients (extrapulmonary source of infection, with severe sepsis or septic shock, n = 11; postextracorporeal circulation with systemic inflammatory response syndrome and/or shock, n = 9; acute pancreatitis, n = 2; subarachnoid and intracerebral hemorrhage, n = 1). One hundred and forty-six patients survived and 97 died, for a case-fatality rate of 40%. The mean number of days of MV for all patients was 19 ± 24, with a median of 9 d; for survivors, the mean number of days of MV was 17 ± 22, with a median of 9.5 d.

Incidence

During the study period, 343 episodes of suspected bacterial lung infection were evaluated with fiberoptic bronchoscopy. However, on the basis of the criteria previously defined, microbiologically provable VAP was documented in only 84 patients, for a total of 106 episodes (14 patients had two episodes and four patients had three episodes). Overall, there was a high level of agreement, both qualitatively and quantitatively, among the three techniques used to establish the diagnosis of VAP. Among the 98 episodes in which both PSB and BAL were performed to ascertain the diagnosis of VAP, only three cases met only one of the three preset criteria (i.e.,  10³ cfu/ml for PSB cultures,  10⁴ cfu/ml for BALF cultures, and  5% of infected BAL cells for microscopy), 49 cases met two criteria, and 46 cases met all three criteria. For the eight episodes in which only BAL was performed because of severe (< 50,000/ mm³) thrombocytopenia, both BALF cultures and microscopy were positive in six cases and only BALF cultures were positive in two cases. Therefore, direct BALF examination detected  5% of infected cells in 77 (73%) of the 106 episodes of VAP.

Thirty-one of the 56 patients with ARDS (55%) developed microbiologically provable nosocomial pneumonia for a total of 41 episodes, as compared with only 53 (28%) of the 187 patients without ARDS for a total of 65 episodes (p = 0.0005). The cumulative hazard function for nosocomial pneumonia in the two groups of patients, using only the first episode of pulmonary infection for analysis, is shown in Figure 1. The actuarial risk of pneumonia during the first 30 d of MV was 14% at 10 d and 58% at 20 d in patients with ARDS, as compared with 24% and 42%, respectively, in patients without ARDS (p = NS).

View larger version (19K): [in this window] [in a new window]   Figure 1.   Kaplan-Meier estimates of the cumulative probability of developing VAP as a function of the number of days on MV for patients with or without ARDS.

Of the 237 episodes for which bronchoscopic results did not confirm the diagnosis of VAP, 81 occurred in 68 of the 84 patients with VAP, either before they developed pneumonia (n = 54 episodes) or after their first episode of lung infection (n = 27 episodes), and 156 episodes occurred in 111 of the 159 patients in whom VAP was never diagnosed. Interestingly, quantitative cultures of BALF revealed low-grade (BALF culture results  10¹ cfu/ml but < 10⁴ cfu/ml) lower-respiratory-tract colonization in 34 of the 54 "negative" episodes that preceded microbiologically confirmed VAP. However, because BALF culture results also demonstrated low-grade lower-respiratory-tract colonization in 95 (61%) of the 156 "negative" episodes that occurred in patients in whom VAP was never diagnosed, only 34 of the 129 episodes in which a low-grade lower-respiratory-tract colonization was demonstrated later evolved to VAP, for a positive predictive value of only 0.26. Restricting the analysis to the 56 patients with ARDS did not improve these results. Only 18 of the 32 "negative" episodes in ARDS patients in which BALF grew  10¹ cfu/ml but < 10⁴ cfu/ml evolved to VAP within 1 to 30 d (mean delay: 10.5 ± 6.5 d), for a positive predictive value of 0.56.

Microbiology of Infection

A total of 208 microorganisms were cultured at significant concentrations from PSB and/or BALF specimens during the 106 episodes of VAP. As indicated in Table 2, the organisms most frequently isolated from ARDS patients were methicillin-resistant Staphylococcus aureus (23%), followed by non-fermenting, gram-negative bacilli (Pseudomonas aeruginosa, Acinetobacter baumannii, and Stenotrophomonas maltophilia) (21%) and Enterobacteriaceae (21%). Sixty-one (58%) of the 106 episodes of VAP were polymicrobial, of which 55% and 60%, respectively, occurred in patients with and without ARDS (p = NS).

When the distribution patterns of microorganisms responsible for infection in patients with and without ARDS were compared, a higher percentage of methicillin-resistant S. aureus was evidenced in the former group. Whereas methicillin-resistant S. aureus represented only 8% of the total number of causative pathogens in patients without ARDS, it represented 23% in patients with ARDS (p = 0.01). Forty-four percent (18 of 41) of all episodes of VAP in patients with ARDS included at least one methicillin-resistant strain of S. aureus, versus only 17% of episodes in patients without ARDS (p = 0.01).

When the analysis was restricted to the first episodes of nosocomial pneumonia, the same distribution pattern of microorganisms responsible for infection was observed, with a still higher incidence of infection caused by methicillin-resistant S. aureus in patients with ARDS (data not shown). However, this difference disappeared completely when patients with ARDS were compared only with non-ARDS patients receiving antibiotics who developed VAP after the first week of MV. As shown in Table 3, the same distribution pattern of microbes was then observed for patients with and those without ARDS.

Clinical Features of VAP

The clinical characteristics of the first VAP episode in the 84 patients who developed one or more episodes are summarized in Tables 1 and 4. Patients with and without ARDS who developed VAP appeared to be very similar with regard to age, severity of underlying illness(es), location before admission to our ICU, and most clinical features at the time of onset of pneumonia, except for lower Pa_O2/FI_O2 ratios and higher radiologic scores, which are expected in ARDS patients. At the time of bronchoscopy, ARDS patients in whom the diagnosis of nosocomial pneumonia was confirmed also showed no significant differences from those in whom this diagnosis was excluded in temperature (mean: 38.4 ± 1.3° C versus 38.3 ± 1.3° C, respectively), white blood cell (WBC) count (mean: 17.0 ± 8.5 × 10³/ml versus 16.4 ± 7.9 × 10³/ml), Pa_O2/FI_O2 ratio (mean: 158 ± 59 mm Hg versus 155 ± 60 mm Hg), or radiologic score (mean: 7.1 ± 2.7 versus 7.8 ± 2.7).

The first episode of microbiologically provable VAP occurred much later after the onset of MV in patients with ARDS than in patients without ARDS. Whereas more bronchoscopies were performed in ARDS patients during the first week of MV than in other patients (mean: 1.4 ± 0.6 versus 0.7 ± 0.6 per patient, respectively; p < 0.001), only 10% of the episodes of VAP in ARDS patients occurred before Day 7 of MV, as compared with 40% of such episodes in patients without ARDS (p = 0.005).

All except two patients with ARDS who developed VAP (94%) were receiving antimicrobial treatment before the onset of pneumonia, as compared with 35 (66%) patients without ARDS (p = 0.004). Not only the frequency of prior antibiotic use, but also the types of antimicrobial agents administered before the onset of VAP differed between patients with and without ARDS. Whereas only 36% of patients without ARDS had received a broad-spectrum antimicrobial agent prior to the onset of pneumonia, 61% of ARDS patients were treated with such an antibiotic agent (p = 0.04). Furthermore, the total number of antibiotic classes previously given was significantly higher in ARDS patients than in non-ARDS patients (1.9 ± 1.0 versus 1.3 ± 1.2; p = 0.03).

Clinical Outcomes

Thirty-four (61%) of the 56 ARDS patients in the study died in the ICU, as compared with 63 (34%) of the 187 patients without ARDS (p = 0.005). Although the complication of microbiologically provable pneumonia in patients without ARDS was clearly associated with a greater mortality than in the nonpneumonia population (47% versus 28%; p = 0.001), the occurrence of VAP did not appear to markedly influence overall survival in patients with ARDS: 16 of the 31 (52%) ARDS patients who developed VAP died, as compared with 18 of the 25 (72%) ARDS patients who did not develop VAP (Table 1). However, of the 18 ARDS patients who died without having developed microbiologically provable VAP, 10 died of multiple organ failure with persistent respiratory failure very early during the course of the syndrome, before the end of the first week of MV, whereas only one of the 16 ARDS patients who acquired VAP died during the same period (p = 0.003). The median time from diagnosis of VAP to death was 5.5 d (range: 1 to 113 d) and 7.0 d (range 1 to 85 d) in nonsurviving patients with and without ARDS, respectively (p = NS).

As indicated in Table 1, the total duration of MV was significantly longer for patients who developed microbiologically provable VAP than for those who did not, regardless of whether or not they had ARDS. The mean durations of MV were 34 and 33 d for VAP patients with and without ARDS, respectively, as opposed to 17 and 10 d for patients without VAP (p = 0.004 and p < 0.0001 for each comparison, respectively).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

To ascertain the epidemiologic and microbial etiologic aspects of VAP in patients with severe ARDS, we undertook a prospective study of 243 consecutive ICU patients who required MV for  48 h, of whom 56 developed ARDS as defined by a Murray LIS > 2.5, and for whom strict bronchoscopic criteria were applied to diagnose pneumonia. Our results indicate that the incidence of microbiologically provable nosocomial pneumonia was particularly high in patients with ARDS, being nearly 2-fold higher than in other ventilated patients, but that such pneumonia occurred later in the course of the syndrome, usually after the first week of MV. Although infection was frequently caused by methicillin-resistant S. aureus and other multiresistant microorganisms, the distribution pattern of infecting organisms in patients with ARDS was very similar to that observed in patients without ARDS who required prolonged MV and who had been receiving antimicrobial treatment, probably reflecting the different use of antibiotics from the onset of MV in these two subsets of patients, and thus the different kinetics of infection. Additionally, no significant differences in survival were detected in ARDS patients with VAP versus those without VAP although the total duration of MV was longer for ARDS and non-ARDS patients who developed microbiologically provable pneumonia than for those who did not.

The diagnosis of pneumonia in patients on MV, and isolation of the causative agents of these patients' pneumonia, are often difficult. To overcome these difficulties, we chose in the present study to systematically perform fiberoptic bronchoscopy, with use of PSB and BAL for each patient clinically suspected of having developed VAP, taking great care to obtain specimens of distal pulmonary secretions before any modifications could be made of existing antimicrobial treatment, if any. Although we cannot exclude the possibility that pneumonia was incorrectly diagnosed in some patients included in our study, it is unlikely that many cases of true pneumonia could have been missed by our diagnostic protocol. Indeed, a heightened clinical suspicion of VAP was maintained throughout the study period for all enrolled patients, and specifically those with ARDS, from whom bronchoscopic samples were obtained as soon as they became febrile and/or deteriorated clinically, even when no progression of lung infiltration could be ascertained. Furthermore, respiratory-tract specimens were obtained from ARDS and non-ARDS patients before any modification of their antibiotic regimens was introduced for any other reason (e.g., surgical-site infection). Additionally, because each of the bronchoscopic techniques used in the study may give rise to a few false-negative results when used alone, we chose to base our diagnostic classification on a protocol combining the results of direct microscopic examination of cells recovered by BAL with the results obtained using quantitative culture of both PSB and BALF specimens. Logically, combining the three techniques should improve and strengthen overall diagnostic accuracy, because false-negative results would probably not occur in all three assays for the same patient (11, 12).

On the other hand, it can be argued that pneumonia was overdiagnosed in our patients, because bronchoscopic techniques may give rise to a few false-positive results despite the universal recognition of these methods as highly specific. However, we think that this hypothesis is unlikely for the following reasons: First, as indicated in Table 4, most patients who developed VAP had, at the time of onset of infection, clinical features highly compatible with pneumonia. Second, a high level of agreement was found among the three techniques used to establish the diagnosis of VAP, both qualitatively and quantitatively: for 47% of episodes, PSB cultures, BALF cultures, and direct BALF examination were positive; for 50% of episodes at least two criteria were positive; and, for only 3% of episodes was only one criterion positive. Moreover, epidemiologic studies (22, 23) of patients who required MV for  48 h have found data concerning the rate of nosocomial pneumonia very similar to ours for patients without ARDS.

The incidence of objectively diagnosed pneumonia in our investigation of patients with ARDS (55%) is comparable to that previously reported in some clinical or autopsy studies, but considerably higher than that reported by Sutherland and colleagues, which was only 15%, despite their similar use of invasive bronchoscopic techniques to diagnose VAP (9). Several possibilities, such as a population with less severe ARDS, a different spectrum of etiologies for the lung disease, and/or a better global strategy for preventing nosocomial infections in Sutherland and colleagues' patients as compared with ours, could explain these divergent results. However, the most probable explanation is that Sutherland and colleagues did not account for the potentially confounding role of antibiotics given during their patients' ICU stays on the culture results with their bronchoscopic samples. Of the 201 BALs performed, 179 were done in the presence of broad-spectrum antibiotics and, most importantly, patients were lavaged at predetermined times, on Days 3, 7, 14, and 21 of their ARDS course, rather than at the time of clinically suspected infection and before new antibiotics were administered (9). It is therefore likely that a number of superinfections went unrecognized in their study, because respiratory-tract samples were in fact obtained during the recovery phase of the infection, at a time when antimicrobial therapy and lung antimicrobial defenses might have been successful in suppressing microbial growth in distal pulmonary secretions. It should be noted that Delclaux and coworkers, who also used bronchoscopic techniques to diagnose VAP in a large series of patients with ARDS, recently reported a 60% incidence of acquired pneumonia (24), thus confirming our data.

Our finding of a higher incidence of microbiologically provable VAP in patients with ARDS than in other populations of mechanically ventilated patients was not unexpected. Several studies have clearly shown that alveolar macrophages and polymorphonuclear cells retrieved from the lungs of patients with ARDS have impaired phagocytic function and/or a lower capacity to express maximal activity after ex vivo stimulation by bacterial products than do corresponding cells from normal subjects, which could explain why these patients are at high risk of developing pulmonary infection (25, 26). However, the actuarial risk of pneumonia after 30 d of MV did not differ significantly between patients with and without ARDS. Therefore, the higher incidence of VAP observed in ARDS patients in our study was essentially the result of their need for a much longer duration of MV than that of other patients, thereby increasing the time during which they were at risk for developing VAP.

It is noteworthy that 44% of all episodes of VAP in the present series of patients with ARDS were caused by or included at least one methicillin-resistant strain of S. aureus, and that 29% included a non-fermenting gram-negative bacillus, such as P. aeruginosa, S. maltophilia, or A. baumannii. Indeed, our ARDS patients who developed microbiologically provable VAP accumulated predisposing factors for an increased risk of infection caused by multiresistant pathogens, as demonstrated by several previous studies (27, 28): 90% of infections occurred after the first week of MV, all patients except two were receiving antimicrobial therapy, and 61% had received broad-spectrum antibiotics during the 15 d preceding the onset of VAP. It must be emphasized that the distribution pattern of microorganisms responsible for infection in our ARDS patients was the same as that observed for non-ARDS patients who required prolonged (> 7 d) MV and who had been receiving antibiotics before the onset of infection, thereby further highlighting the observation that the high incidence of multiresistant pathogens observed in ARDS patients was not so much the consequence of the peculiar type of the lung disease as a reflection of the different kinetics of infection and different use of antimicrobial agents in the two populations. Clearly, the early administration of broad-spectrum antibiotics to ARDS patient could have delayed the onset of VAP, preventing early-onset episodes and favoring the occurrence of late-onset episodes caused by multiresistant pathogens. Certain limitations to the microbiologic aspects of this study, however, merit attention. First, the microbiologic features associated with pneumonia in this study reflect the type of ICU patients admitted to our unit, and may therefore not be applicable to all patients receiving MV; and second, the prevalence of methicillin-resistant S. aureus is particularly high in France, which could explain the high rate of such infections in our patients (29).

In agreement with the findings in many previous studies, patients without ARDS who developed microbiologically provable VAP in our study had nearly a 2-fold greater mortality rate than nonpneumonia patients (15, 30, 32). We were unable to document the same increase in mortality for ARDS patients, although many cases of VAP in this group of patients were due to multiresistant, difficult-to-treat pathogens, confirming previous data reported by other investigators (33, 34). In fact, patients with ARDS who did not develop VAP, but who nevertheless died, did so earlier than other ARDS patients. Although most pneumonias in ARDS patients occurred after the first week of MV, 10 of the 18 patients with ARDS who died without having developed VAP died before Day 7 of MV, and therefore had little opportunity to develop nosocomial pneumonia. Interestingly, all patients with ARDS who died early in the course of their disease underwent fiberoptic bronchoscopy with PSB and/or BAL before their death. It is therefore unlikely that the diagnosis of VAP was missed in these patients and that they were misclassified as not having VAP. The total duration of MV was significantly longer for patients with microbiologically provable VAP, both those with and without ARDS, than for other patients who did not acquire VAP, underlining the excess morbidity associated with this complication, as previously reported (15, 30, 31).

In conclusion, the findings of this study indicate that microbiologically provable nosocomial pneumonia, as diagnosed with invasive bronchoscopic techniques, is really a common complication of ARDS and occurs far more often in this group of patients than in other ventilated patients. The most probable explanation for this high incidence of VAP in this population is their need for a much longer duration of MV, since the actuarial rate of pneumonia for patients with and without ARDS is similar. Because ARDS patients are often treated very early with broad-spectrum antibiotics, infection is frequently delayed until after the first week of MV, and is then caused mainly by methicillin-resistant S. aureus and other multiresistant organisms.