INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The diagnosis of superimposed pneumonia in patients receiving mechanical ventilation because of adult respiratory distress syndrome (ARDS) remains a difficult challenge. Indeed, the clinical criteria commonly used to diagnose ventilator- associated pneumonia (VAP) (i.e., fever, leukocytosis, purulent sputum, and new or persistent radiographic infiltrate) are almost always present in such patients or at least during a large part of their stay in the intensive care unit (ICU). Specifically, radiographic changes suggesting pneumonia may be extremely difficult to detect in patients with ARDS (1, 2). Likewise, most patients with ARDS have a systemic inflammatory response syndrome (3) and fever and high leukocytosis are common, even in the absence of infection (4); moreover, infection itself is often associated with the onset of the syndrome. Conversely, clinical studies and pathologic examination of the lungs of patients who had died from ARDS have suggested that nosocomial pulmonary infection is frequent and often associated with the development of multiple organ failure and eventual death (5, 6); many of these patients had pulmonary infection on pathologic examination of the lungs that had apparently not been suspected before death (4, 5).

Since the clinical criteria of VAP lack both sensitivity and specificity in this setting, microbiologic information achieves major importance in the diagnostic strategy of infection. Because of the widespread colonization of the tracheobronchial tree in patients on prolonged mechanical ventilation (7), various sampling techniques or even a combination of techniques have been advocated to retrieve uncontaminated lower respiratory tract (LRT) secretions, including protected brushing (8, 9), protected catheter (10), and bronchoalveolar lavage, whether protected or not (11, 12). Using these techniques, a diagnosis of pneumonia is usually accepted when samples yield a bacterial growth above a predefined threshold.

A recent study by Sutherland and colleagues (2), using quantitative culture of BAL fluid as a diagnostic tool, found however an incidence of VAP of only 15% in 105 patients with ARDS and therefore concluded that VAP was not more common in ARDS than in other circumstances, so that its incidence may have been previously overestimated. A major problem with the interpretation of cultures of respiratory tract samples, however, is the confounding role of antibiotics that are administered at time of sampling. Contrarily to antibiotics that have been administered unchanged for several days before sampling (12, 13), changes in therapy made within a short time of samplings (i.e., within 48 to 72 h) are very likely to interfere with the recovery of organisms from subsequent samples. In the above-mentioned study (2), antibiotics administered and the delay between changes in antibiotics and sampling were not taken into account in the analysis of culture results of respiratory tract samples.

We therefore undertook this study to investigate further the incidence of pulmonary superinfection during severe ARDS and the sequence of colonization to infection, as determined by quantitative culture of serial single-sheathed plugged telescopic catheter (PTC) samples (10) obtained blindly during the course of ARDS, and bronchoscopic bronchoalveolar lavage (BAL) when indicated. We used PTC to sample respiratory tract secretions, because this technique has high sensitivity and acceptable diagnostic specificity and because it can be easily performed repeatedly without fibroscopy (i.e., "blindly"), even in patients with severe hypoxemia. Whenever superimposed VAP was suspected from clinical and microbiologic information, new samples were taken before any change in ongoing antimicrobial therapy.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This study was conducted from January 1994 to July 1996 in the medical intensive care unit (MICU) of Henri Mondor Hospital, a tertiary care referral hospital with 1,028 beds and eight ICUs. Patients with severe ARDS admitted from the community or transferred to the MICU from other wards or ICUs were prospectively identified and evaluated daily in the MICU for presence of infection. Severe ARDS (14) was defined as: (1) generalized pulmonary infiltrates involving all four lung quadrants on chest roentgenogram; (2) a Pa_O2/FI_O2 ratio  150 mm Hg, persisting at least 24 h, regardless of the level of positive end-expiratory pressure at entry into the study; (3) a pulmonary capillary wedge pressure < 18 mm Hg; (4) a recognized cause for the development of ARDS. Causes of ARDS were classified as direct or indirect lung injury. A direct lung injury was defined as a primary insult to the lung such as that occurring with pneumonia or aspiration of gastric content; indirect lung injury was defined as an extrapulmonary insult such as that occurring with an extrapulmonary source of infection, systemic inflammation, or shock.

Study Design and Definitions

Surveillance samples for microbiologic examination and cultures were obtained every 48 to 72 h, as part of routine surveillance of LRT colonization, until weaning from mechanical ventilation or death. Routine samples consisted of PTC taken blindly via the endotracheal tube (10). Whenever possible, a repeated PTC sample and/or a BAL were performed, both via fibroscopy, to obtain further confirmation of infection, when routine PTC samples yielded bacterial growth in a concentration at or above the predefined threshold of 10³ colony-forming units (cfu)/ml by quantitative culture. A repeated sample was similarly obtained when clinical features consistent with pneumonia were met, before any change in antibiotic therapy and/or when a change was seen on chest roentgenogram (new infiltrate or change in preexisting infiltrates), to allow bacterial sampling in the area of the new infiltrate.

The clinical criteria for infection were those previously used in a similar study (2). A diagnosis of probable or possible pneumonia was made if the patient met at least three of the four following criteria: (1) fever (defined as a rectal temperature  38.3° C); (2) leukocytosis (total white blood cell count > 10,000/mm³); (3) increase in volume or new onset of purulent sputum; (4) development of a focal radiographic infiltrate.

Antimicrobial therapy could be initiated or modified by the attending physician, whenever a PTC yielded organisms in significant growth (i.e., above the threshold value of 10³ cfu/ml) was obtained and usual clinical features consistent with probable or possible pneumonia were met; as stated above, physicians were requested to obtain a repeated PTC or BAL sample in this situation before starting new antibiotics; however, empiric therapy could be started before culture results were available in case of severe sepsis or septic shock.

For the interpretation of PTC culture results, we used the following definitions: lower respiratory tract colonization was defined as a PTC culture yielding < 10³ cfu/ml with or without clinical criteria of VAP, or one PTC culture  10³ cfu/ml, in the absence of clinical criteria for possible or probable VAP; pulmonary infection was defined as a PTC culture yielding  10³ cfu/ml in the presence of clinical criteria of possible or probable pneumonia. In this situation, a repeated PTC sample was taken after receipt of culture results and before starting new antibiotic therapy for VAP; whenever possible, a BAL was also obtained in case of non-life-threatening hypoxemia (Pa_O2/FI_O2 > 60 mm Hg) to confirm the bacteriologic diagnosis of VAP using PTC culture, BAL culture, and examination of the proportion of alveolar cells containing intracellular bacteria on May Grünwald-Giemsa-stained lavage slides (12). BAL was performed using three 50-ml aliquots of sterile 0.9% saline. In case of a new radiographic infiltrate, BAL was performed in the subsegment involved after performing PTC; in other cases, BAL was performed in the right middle lobe or lingula and PTC was performed in the postero-basal subsegment. The threshold for quantitative culture of BAL fluid used for diagnosing pneumonia was  10⁴ cfu/ml of one or more bacteria in BAL fluid. Direct examination of lavage slides was considered positive when  2% cells with intracellular bacteria were seen (15).

Clinical Features of VAP and Outcome

As the usual clinical criteria for VAP were expected to have low predictive value, we examined whether criteria indicating a significant worsening of the clinical status of the patient could be better predictors of the occurrence of VAP, as defined above according to results of microbiologic samplings. The following changes, recorded prospectively and analyzed for the 48 h preceding a positive sample, were arbitrarily defined as a significant change in the clinical condition: (1) a change in temperature of  1° C above baseline; (2) an increase in leukocytosis of  20% above baseline; (3) an increase in volume and/ or purulence of sputum (semi-quantitatively defined); (4) a new radiographic infiltrate; (5) a decrease in Pa_O2/FI_O2 ratio  20%; (6) the occurrence of septic shock.

The outcome of antimicrobial therapy of VAP was evaluated as follows: (1) clinical and bacteriologic cure was defined as an improvement of clinical features of VAP and sepsis, associated with eradication of the causative organism(s) on at least two subsequent consecutive PTC samples; (2) bacteriologic cure only was defined as repeated negative PTC samples in the absence of clinical improvement; a treatment failure was defined as: (3) no clinical improvement or death with persistence of the original organisms or (4) superinfection with growth of new organisms in significant concentration on subsequent PTC samples.

Statistics

Data are reported using mean values and SD for comparisons between ARDS patients with and without VAP. Differences between groups used Student's t test, and a value of p < 0.05 was considered statistically significant.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients

Thirty consecutive patients with severe ARDS, including 21 male (70%) and nine female patients, were studied (Table 1). ARDS was caused by a direct lung injury in 18 patients (aspiration, n = 7; community-acquired pneumonia, n = 7; miscellaneous, n = 4) and was secondary to an indirect lung injury in 12 patients (severe sepsis or septic shock, n = 8; miscellaneous, n = 4); six patients had severe immunosuppression (AIDS, n = 2; leukemia, n = 2; liver transplantation, n = 1; cancer chemotherapy, n = 1). All patients received antimicrobial therapy during the course of ARDS. All patients received mechanical ventilation for at least 5 d (range, 5 to 49 d), and patients who developed VAP remained on mechanical ventilation significantly longer (mean ± SD, 23.6 ± 13 d) than patients without VAP (12.5 ± 13 d; p < 0.05). The overall mortality in this selected group of patients with severe ARDS was 83%, and only five patients survived.

Incidence and Diagnosis of VAP

Twenty-four episodes of VAP were diagnosed in 18 patients (four patients had two episodes, and one had three episodes), according to culture results of serial PTC and BAL. Thus, the incidence of VAP in our patients with severe ARDS was 60%, and its incidence density was 4.2/100 d of mechanical ventilation. In 10 of 24 (42%) confirmed VAP episodes, a BAL procedure could be performed before initiating antimicrobial therapy and within 48 h of the first positive PTC sample. In nine of these 10 cases, the same organism(s) as found on PTC culture grew in BAL fluid in a concentration 10⁴ cfu/ml. Intracellular organisms (mean, 16 ± 22% of white blood cells; range, 2 to 65%) were found in seven of 10 BAL fluid samples, in agreement with the positive PTC. In only one patient with a positive PTC, BAL fluid culture grew the same organism as PTC but below the threshold of 10⁴ cfu/ml. Overall, there was a good agreement between PTC and BAL microbiologic criteria in patients evaluable in this respect. The organisms recovered from all episodes of VAP are listed in Table 2. Acinetobacter and Pseudomonas species were the most commonly identified gram-negative organisms, and Staphylococcus aureus was the predominant gram-positive organism.

Surveillance Samples and LRT Colonization

Two to 20 serial PTC samplings (a mean of 6.8 ± 4.5 per patient) were obtained in each patient for surveillance cultures, and a total of 204 PTC were obtained. Eighteen episodes of LRT colonization occurred in 14 of 30 (47%) patients. Colonization was associated with PTC growing < 10³ cfu/ml in 10 episodes or growing  10³ cfu/ml but with less than three clinical criteria for infection in eight episodes, including four having first a PTC growing < 10³ cfu/ml. Sixteen of 18 colonization episodes evolved to pneumonia within 2 to 6 d (mean delay, 4.1 ± 1.2 d); only two episodes of colonization (one each with Pseudomonas aeruginosa and Staphylococcus epidermidis, at a growth of < 10³ cfu/ml) did not evolve to VAP. The positive predictive value of LRT colonization for infection was therefore 0.89 in our series of patients with severe ARDS. However, LRT colonization preceded microbiologically confirmed VAP in only 16 of 24 (67%) VAP episodes; the negative predictive value for VAP of colonization was therefore only 0.60. In all patients with colonization followed by confirmed infection, as defined above, identical organisms by species identification and antimicrobial susceptibility profilewere found on two to three successive PTC samples, and no other organisms were recovered. A similar pattern of organisms causing infection was found whether or not LRT colonization preceded infection (Table 2).

Antibiotics prior to VAP

Each patient received antibiotics during his/her course in the ICU; however, such therapy had been introduced or modified a mean of 10 ± 3.6 d before a diagnosis of VAP was made (range, 4 to 15 d), and one patient had not been receiving antibiotics for 8 d at the time of positive PTC sample. In no patient were new antibiotics introduced before attempting to confirm a clinical suspicion of pneumonia. Antibiotics administered during the course of ARDS and prior to the onset of acquired VAP were similar in patients with or without VAP: eight of 18 (44%) and five of 13 (38%), respectively, received amoxicillin alone or combined with clavulanic acid as therapy for proved or suspected community-acquired pulmonary infection and the others received broad-spectrum combination therapy for various other, mostly hospital-acquired infections.

Clinical Features of VAP

The clinical characteristics of the first VAP episode in the 18 patients acquiring one or more episodes of infection are summarized in Table 3. On average, VAP occurred 9.8 ± 5.7 d (range, 4 to 28 d) after onset of ARDS. The occurrence of VAP was associated with some degree of worsening of the patient's condition (Figure 1). The occurrence of the first VAP episode was associated with worsening of hypoxemia in seven patients, and new respiratory signs (increase in sputum, new infiltrate, or worsening of hypoxemia) were noted in 50% of VAP episodes. A new radiographic infiltrate was observed in seven patients, of which only four were associated with a microbiologically confirmed VAP; other episodes of new radiographic infiltrates were attributed to the occurrence of segmental atelectasis. Thus, the positive predictive value of a new radiographic infiltrate for diagnosing pneumonia was only 0.57. Similarly, a sustained decrease of the Pa_O2/FI_O2 ratio was recorded 15 times, of which nine were associated with the occurrence of VAP; thus, the worsening of hypoxemia had a positive predictive value of only 0.60 in our series. Although the predictive value of individual signs was rather low, all VAP episodes were associated with at least one worsening clinical sign or symptom (Figure 1).

View larger version (13K): [in this window] [in a new window]   Figure 1.   Frequency of occurrence of six worsening clinical criteria recorded within 48 h of diagnosis in 24 episodes of VAP in 18 patients with severe ARDS. Fever: increase in temperature of  1° C above baseline; leukocytosis: increase in blood leukocyte count of 20% above baseline; hemodynamic: occurrence of septic shock; sputum: increase in volume and/or purulence of endotracheal secretions; infiltrate: new infiltrate; hypoxemia: decrease in PaO2/FIO2 ratio of  20% from baseline.

Outcome of Therapy

Clinical and microbiologic cure with antimicrobial therapy was obtained in only eight of 18 (44%) patients for the first episode of pneumonia and clinical and microbiologic treatment failure occurred in six patients (Table 3). Although VAP appeared to contribute directly to death in only four patients all patients (eight of eight) who failed initial therapy died, as well as four of four patients (one of these patients died after a second episode of VAP) who had bacteriologic cure only, whereas nine of 11 patients who had both clinical and bacteriologic cure survived. However, the occurrence of VAP did not appear to markedly influence overall survival in this selected group of severely ill patients: 14 of 18 (78%) patients acquiring VAP died, as compared with 11 of 12 (92%) patients without VAP (Table 1); 68% of deaths were associated with persisting respiratory failure, often together with other organ failure (mean organ system failure score of 2.0 ± 0.5 on admission and of 2.7 ± 1.0 on the day preceding death).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This study was designed to assess the incidence and characteristics of lower airways colonization and VAP in a series of patients having severe ARDS. Using repeated microbiologic assessment of colonization and infection of the lower airways, we found a high incidence of VAP, preceded by colonization in about two thirds of episodes.

As clinical criteria of VAP, including the radiologic criteria, are of especially low diagnostic value during ARDS (1, 4, 9), microbiologic criteria tend to gain preeminence on other diagnostic information in this setting. The use of quantitative culture techniques on PTC and/or BAL specimens obtained at predefined intervals allowed us to prospectively and accurately determine the incidence and sequence of LRT colonization to infection in patients with ARDS. We found a 60% incidence of VAP in patients receiving mechanical ventilation for severe ARDS. These findings are at variance with those of another recent prospective study by Sutherland and colleagues (2) using quantitative culture of BAL specimens that were to be taken at predetermined intervals; in that study, the incidence of VAP in patients with ARDS was only 15%. Several explanations could account for the divergent results between that study and our study: first, patients with ARDS studied by Sutherland and colleagues were less severely ill than in our study, and it is likely that more severely ill patients such as ours are at higher risk of superinfection (5, 16); second, only one BAL sample was obtained during the course of ARDS in about half of their patients, although the duration of mechanical ventilation averaged 15 d; third, the investigators did not account for the potential confounding role of antibiotics administered during the ICU stay of their patients on the culture results of respiratory tract samples. It is therefore likely that a number of occult superinfections went undetected by Sutherland and colleagues because of the lack of samples and/or because of concurrent changes in antibiotic therapy. In our study, every attempt was made to obtain respiratory tract samplings before any change in antibiotics administered: although almost all patients were receiving antibiotics at time of sampling and suspicion of pneumonia, no change in prior antibiotic therapy was made before a confirmatory sample was taken, even though empiric therapy could be instituted before receiving culture results. Such a diagnostic strategy was made possible through the repeated use of blind PTC samplings, a technique that can be easily performed any time, even in thrombocytopenic and/or severely hypoxemic patients, and that can be performed without fibroscopy (10).

It could be argued that we have overdiagnosed VAP in our patients. Indeed, PTC is viewed as highly sensitive but possibly less specific than other protected sampling techniques used for diagnosing VAP (10). However, we believe this hypothesis is unlikely because of the following: (1) when both BAL and PTC could be performed to confirm VAP, the results of quantitative cultures of both specimens were usually concordant (nine of 10 episodes), both qualitatively and quantitatively, thus increasing the number of microbiologic criteria for pneumonia (PTC and BAL quantitative cultures, and microscopic examination of BAL fluid and intracellular bacteria); (2) all patients classified as having VAP had at least three of the usual clinical criteria for VAP and demonstrated at least one criterion for worsening of their clinical condition, with new respiratory symptoms in one half of episodes; (3) studies in ventilated patients, including ours, have demonstrated a relatively low rate of false-positives of protected samples which does not exceed 20% (9, 10); (4) data derived from continuous surveillance of nosocomial pneumonia for 1995 to 1996 in our ICU, using the same sampling technique and microbiologic criteria, indicate that the overall incidence of VAP is 17% and 32%, respectively, among all patients receiving mechanical ventilation (n = 738) and in those receiving mechanical ventilation for  48 h (n = 415), irrespective of the reason for mechanical ventilation (unpublished data); this is in accordance with a 27% median incidence rate found by George (17) when combining 22 epidemiologic studies of pneumonia in mechanically ventilated patients and with the reported incidence ranging from 9% to 27% in studies using quantitative cultures of protected specimens to diagnose VAP in patients undergoing mechanical ventilation for > 48 h (9, 18, 19).

Although rates of VAP may be appreciated differently in series of ARDS from different institutions, a high incidence of VAP in patients with severe ARDS is not unexpected. A defect in host leukocyte-macrophage functions could account in part for a high risk of superinfection in ARDS. A few studies have examined alveolar polymorphonuclear (PMN) functions during ARDS and have found the phagocytic activity of alveolar PMN impaired as compared with vascular PMN from the same patients, or a low capacity of ex vivo stimulation of alveolar PMN by bacterial peptides, which could participate to the pathogenesis of VAP (20, 21). However, alveolar PMN functions do not seem to be related to the severity of respiratory failure, as suggested recently (22). On the other hand, the increased frequency of VAP during ARDS could be directly related to the global severity of illness and not only to the degree of respiratory failure itself. Along this line, recent multivariate analyses of risk factors for pneumonia have suggested that the occurrence of VAP is associated with the severity of illness, whether assessed by a global severity score or the number of organ failures rather than with the severity of hypoxemia itself (16, 23).

It is noteworthy that VAP was of relatively late occurrence during mechanical ventilation in our patients. However, all patients received more or less broad-spectrum antibiotic therapy early in the course of ARDS; this fact could partially explain the absence of early-onset VAP in our population. Most such pneumonias are due to common pathogens that are usually susceptible to many antibiotics, and early antibiotics administered could have delayed the onset of VAP by preventing early-onset episodes and favoring the occurrence of late-onset pneumonias due to more resistant high-risk organisms (24), such as those recorded in our series.

The high mortality in our series is likely explained by the selection of a group of patients presenting with the most severe form of ARDS. The population we studied is known to be at very high risk of mortality and would be more homogenous than all patients with ARDS, as defined by the recent European-American consensus conference (25). Indeed, a large study of ARDS (14) has shown that the Pa_O2/FI_O2 ratio after 24 h of treatment clearly identifies two populations of patients with strikingly different mortality rates: patients with a Pa_O2/ FI_O2 ratio < 150 mm Hg are expected to have a mortality rate of about 70%. A substantial fraction of our patients had severe immunodepression and most had associated organ dysfunction, with a mean of two organ failures on admission, both factors known as major determinants of mortality in ARDS (5, 26). In this severely ill group, the occurrence of pneumonia did not appear to increase the risk of death, although many cases were due to high-risk organisms such as P. aeruginosa and Acinetobacter baumannii. Rather, mortality rate tended to be lower in patients with pneumonia as compared with patients without VAP (78% versus 87%). This is in apparent contradiction to recent observations by several investigators (24, 27); however, this observation should be interpreted in light of the duration of exposure to the risk of VAP because of mechanical ventilation. Patients without pneumonia tended to be sicker at admission and died earlier than the others (Table 1); many patients without secondary VAP died before day 10 of mechanical ventilation, whereas VAP occurred a mean of 11.6 ± 6.2 d after onset of mechanical ventilation and about 10 d after onset of severe ARDS.

Our study documents the frequent occurrence of lower airways colonization in patients with ARDS and subsequent infection. Upper airways colonization with hospital-acquired organisms usually occurs within 48 to 72 h of mechanical ventilation, often leading to secondary VAP (28). It should be emphasized that we used protected catheters, not tracheal aspirates, to assess lower airways colonization in this series. Such sampling technique protects from upper airways contamination of the sample and organisms recovered better reflect those present in the lower airways that are more likely to cause infection. Accordingly, LRT colonization rates found in this study are lower than tracheo-bronchial colonization rates usually found using nonprotected tracheal aspirates in mechanically ventilated patients. For example, de Latorre and coworkers (29) found 90 of 100 mechanically ventilated patients colonized at some point during their ICU course, using quantitative cultures of tracheal aspirates, while only 12 were diagnosed as having pneumonia. It is also likely that the widespread use of antibiotics in our population influenced the results, by selecting resistant pathogens causing subsequent infection.

In our study, almost all colonization episodes evolved to infection within 2 to 6 d, but colonization preceded infection in only two thirds of VAP episodes. Therefore, the occurrence of colonization of the LRT with potential pathogens should prompt for careful screening and follow-up for associated criteria of infection and consideration of early therapy. A few other studies have examined the sequence of tracheo-bronchial colonization to VAP. Using quantitative cultures of repeated nonprotected tracheal aspirates, Salata and colleagues (30) found higher bacterial counts (1.4 × 10⁶ cfu/ml) in patients developing infection as compared with patients remaining colonized (1.3 × 10⁴ cfu/ml); however, colony counts of  10⁵ cfu/ml were not specific for infection. In a study of similar design to ours, A'Court and associates (31) studied LRT colonization using quantitative cultures of repeated mini-BAL (20 ml of saline) performed on alternate days; of 150 mechanically ventilated patients, 65 (43%) were diagnosed as having VAP; in the latter patients, a significant increase in bacterial counts (from  10³ to  10⁵ cfu/ml) was noted in the 2 d preceding the diagnosis of VAP. Our results confirm these findings in patients with ARDS and emphasize the high predictive value for infection in this population of LRT colonization as assessed by positive cultures from PTC samples.

Our data suggest that LRT superinfection may be underestimated in patients with ARDS, especially in the most severe group, often receiving multiple antibiotic courses. Given the results of surveillance cultures and outcome of colonization in our patients, earlier therapy may be warranted in such patients, even when the usual threshold for positivity of samples is not reached. Such a strategy in our patients would have resulted in treating by 2 to 6 d earlier 14 of 18 patients eventually diagnosed as having pneumonia by several sampling techniques; conversely, only two patients would have been treated unduly. Several studies using postmortem biopsies in patients who had died from severe acute respiratory failure have shown that pathologic features consistent with pneumonia may be found in patients on antibiotics in association with low bacterial burden in the lungs (15, 32, 33). Such features may represent early stages of infection and may be recognized by increasing growth of pathogens on serial PTC samples, thus allowing earlier diagnosis of infection and therapy.

Although it is unlikely that earlier therapy would have altered the prognosis in this particular series of patients with such high severity of ARDS, it may be asked whether administering antibiotic therapy as soon as LRT colonization occurs, in association with any alteration in respiratory status, such as a decrease in Pa_O2/FI_O2 ratio, may alter the outcome of patients with less severe forms of ARDS. This hypothesis needs further investigations in corresponding groups of patients, weighing the benefits and risks in this population of a potential overuse of antibiotics.