INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Acute respiratory distress syndrome (ARDS) has been a challenging organ-failure condition with a high mortality rate from its first description as a syndrome in 1967. Nosocomial pneumonia is a frequent complication of ARDS. Although the impact of nosocomial pneumonia on the outcome of mechanically ventilated patients remains controversial (1, 2), it may play a key role by worsening hypoxemia and causing sepsis, multiple organ failure, and death. Previous clinical and histologic studies have found that ventilator associated pneumonia (VAP) can affect between 20% and 75% of patients dying of ARDS (3). However, it is difficult to diagnose nosocomial pulmonary infection in patients with ARDS because the clinical criteria (fever, leucocytosis, and purulent tracheal secretions) for VAP are nonspecific, and since all patients with ARDS have bilateral diffuse infiltrates on chest radiographs, radiologic changes may be very difficult to detect. Most previous studies have used the bacteriologic results of tracheal aspiration, a technique with poor specificity because of tracheal colonization during mechanical ventilation, for diagnosing such pulmonary infection. Recently, four groups of investigators studied the incidence of pulmonary superinfection during ARDS and obtained conflicting results for the exact incidence of VAP. All of these previous studies were conducted at single centers, and none examined the potential risk factors for acquisition of VAP during ARDS (7). We therefore undertook this prospective multicenter study to: (1) investigate the incidence of microbiologically documented VAP in a large number of patients with ARDS; (2) compare the mortality and morbidity in ARDS patients with and without VAP; and (3) evaluate risk factors for pneumonia in these patients. The diagnosis of VAP was based on quantitative cultures of samples obtained with the protected specimen brush (PSB), bronchoalveolar lavage (BAL), or plugged telescopic catheter (PTC) techniques via fiberoptic bronchoscopy. Sucralfate was strongly associated with an increased incidence of VAP in this study. This finding seems to contradict widely held opinion (11, 12).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Design and Data Collection

The study was conducted at eight medical and three medical-surgical intensive care units (ICUs) of 10 hospitals in France. The study protocol was approved by the Institutional Review Board of the Société de Réanimation de Langue Franaise. Patient consent was not required, since all procedures were part of routine care. ARDS was prospectively considered to be present if a patient met the criteria of the North American-European consensus conference of showing: (1) diffuse infiltrates visible on a chest radiograph; (2) a ratio of Pa_O2 to the fraction of inspired oxygen (FI_O2) of less than 200, whatever the positive end-expiratory pressure; (3) a pulmonary capillary wedge pressure (Pcwp) < 18 mm Hg in patients with a pulmonary artery catheter, or no clinical or echocardiographic evidence of congestive heart failure in those without; and (4) no other explanation for the foregoing findings. Patients were excluded if the following criteria were met: (1) moribund state (i.e., death expected within 24 h); (2) evidence of left cardiac failure documented by pulmonary artery catheterization or echocardiography; (3) use of selective digestive decontamination; (4) neutropenia of any origin, defined by a granulocyte count of < 500 cells/mm³; and (5) ARDS of more than 72-h duration.

The following variables were recorded on admission to the ICU: age, sex, severity of underlying disease classified by standard definitions according to the MacCabe score (13), and severity of illness evaluated with the Simplified Acute Physiologic Score (SAPS II) (14). Organ failures were assessed with the Organ Dysfunction and/or Infection (ODIN) score (15). The following variables were recorded on inclusion of a patient in the protocol: cause of ARDS, maximal temperature in the preceding 24 h, leukocyte count, SAPS II, ODIN score, arterial blood gas tensions, and plateau airway pressure. The use of antibiotics was noted. Antibiotics such as penicillin G, glycopeptides, and macrolides were considered to be "narrow spectrum" antibiotics. All other antibiotics (amoxicillin-clavulanate, third-generation cephalosporins, fluoroquinolones) were classified as "broad-spectrum" antibiotics. The following potential risk factors for pneumonia were noted until the first episode of VAP, or until weaning or death in patients who did not develop VAP: stress ulcer prophylaxis (omeprazole, sucralfate, H₂-blockers, or antacids), prospectively defined as the use of these agents for at least 48 h; oro- or nasotracheal intubation; tracheostomy; number of reintubations; enteral or parenteral nutrition for at least 48 h; use of corticosteroids ( 0.5 mg/kg/24 h prednisone); sedation (defined as a combination of a benzodiazepine, flunitrazepam  1 mg/h or midazolam  5 mg/h, and fentanyl > 50 µg/h or sufentanyl > 5 µg/h) for at least 2 h; and use of myorelaxants (pancuronium bromide or vecuronium > 1 mg/h) for at least 2 h.

The main parameter studied was the incidence of VAP in ARDS patients. Patients were monitored for the occurrence of pneumonia during mechanical ventilation until death (in patients dying during mechanical ventilation) or until successful weaning (without need for reintubation within the 3 d after extubation). Three other parameters were studied: the impact of pneumonia on survival in the ICU and on the duration of mechanical ventilation; the pathogens recovered from respiratory tract specimens; and risk factors associated with the development of a first episode of pneumonia.

All patients admitted during the study period and who required mechanical ventilation for > 48 h were analyzed to compare the incidence of pneumonia in patients with and without ARDS. The following parameters were studied in non-ARDS patients: age, SAPS II, duration of mechanical ventilation, development of VAP, pathogens involved, and outcome in the ICU. Non-ARDS patients underwent the same fiberoptic bronchoscopic techniques for culture as those performed in ARDS patients. All patients (ARDS and non-ARDS) were kept semirecumbent as much as possible.

Definitions

The presence or absence of pulmonary or extrapulmonary infection and the etiologic agents were noted. A pulmonary infection was considered to be community acquired if it occurred in the community and was present on admission to the hospital, or was considered nosocomial if it manifested more than 48 h after admission to the hospital. VAP was defined as a pneumonia occuring after more than 48 h of mechanical ventilation and for up to 72 h after weaning. Samples for quantitative cultures were used to diagnose pneumonia, and were taken in cases showing new radiographic infiltrates plus at least one of the following criteria: (1) fever (defined as a temperature > 38° C) or hypothermia (defined as a temperature < 36° C); (2) purulent tracheal aspirates; or (3) leukocytosis (total white blood cell count > 10,000/mm³). Samples were also taken for quantitative cultures, even in the absence of the preceding criteria, in the following situations: (1) development of persistent hypotension, defined by a systolic blood pressure of < 90 mm Hg despite fluid resuscitation with 500 ml colloid or the use of a vasopressor to maintain blood pressure (dopamine > 5 µg/kg/min, epinephrine, or norepinephrine), or by the need to increase the dose of catecholamines by  50% for more than 2 h; or (2) unexplained worsening of arterial oxygenation, defined by a 30% decrease in Pa_O2/FI_O2 for more than 2 h. In all cases, specimens were taken before the introduction of new antibiotics. VAP was diagnosed on the basis of the results of fiberoptic bronchoscopic examination with the PSB, BAL, or PTC technique. Quantitative culture criteria used to diagnose pneumonia were  10³ cfu/ml for PSB cultures (16),  10⁴ cfu/ml for BAL cultures (17), or  10³ cfu/ml for PTC cultures (18). The diagnosis of pneumonia was also accepted in the presence of a microorganism in pleural effusions or isolation of the same microrganism (except coagulase-negative staphylococci and Corynebacterium sp.) from at least one blood culture as well as a respiratory tract specimen, but in numbers below the threshold cfu values given earlier, in the absence of other foci of infection. The study physicians were given the choice of starting antibiotics just after the performance of bronchoscopic sampling, or witholding this decision pending culture results.

Statistics

Patients with and without pneumonia were compared through the use of Student's t test for continuous variables, or through contingency table analysis for categorical data. Fisher's exact test was used when expected values were less than 5. Nonparametric estimates of the functions describing continuation of mechanical ventilation versus time for patients with and without pneumonia were plotted according to the Kaplan-Meier method and compared through use of the log-rank test (19). This analysis was first performed on all patients. Data for patients free of pneumonia were terminated at the time of their death or at the end of the protocol (3 d after succesful extubation). The analysis was then repeated, including those patients who survived until weaning. Risk factors for pneumonia were evaluated only for the first episode of pneumonia. These factors were assessed by two means: first by univariate analysis with Bonferroni's correction to account for the multiplicity of comparisons, and second by logistic regression analysis. Variables associated with an uncorrected value of p < 0.05 by univariate analysis were entered into the regression. Data on H₂-blockers were forced into the model, given their importance as a potential risk factor, irrespective of their significance on univariate analysis. Adjusted odds ratios (ORs) and their 95% confidence intervals (CIs) were calculated through standard methods. Data are presented as means ± SD, or as number or percentage of patients. A value of p < 0.05 was considered significant (otherwise indicated in case of multiple comparisons). All analyses were done with the Statview 5 (SAS Institute Inc., Cary, NC) statistical software package.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Population and Incidence and Occurrence of VAP

A total of 878 patients were mechanically ventilated for at least 48 h. Of these, 134 (15%) had ARDS and were enrolled in the study. The demographic and baseline characteristics, comorbidities, and severity scores on admission to the ICU of the ARDS patients with and without VAP were not significantly different. A total of 86 patients (64%) had infections at admission; these infections were of pulmonary origin in 80% of those cases (Table 1). The clinical characteristics and severity scores of the 134 patients at the onset of ARDS are shown in Table 2. VAP was diagnosed in 49 patients (36.5%) with ARDS, with a total of 73 episodes (eight patients had two episodes, five patients had three episodes, and two patients had four episodes). Patients with and without VAP were similar with respect to clinical variables, SAPS II, number of organ failures, gas exchange values, plateau airway presure during mechanical ventilation, previous antibiotic treatment, and etiology of ARDS. The syndrome was caused by pulmonary and extrapulmonary infections in the majority of patients (Table 2). The mean duration of mechanical ventilation before the first episode of VAP was 11.7 ± 11.9 d. The incidence of early-onset pneumonia was low, since only 12 patients (35%) developed pneumonia within the first 5 d. In contrast, 65% of VAP episodes (37 of 49) occurred after Day 5 of mechanical ventilation. A first pneumonia very rarely occurred after 20 d on mechanical ventilation (four of the 49 patients). Table 3 shows the frequency distribution of the first pneumonia according to time. The incidence of microbiologically documented pneumonia during the same period was only 23% (173 of 744 cases) among patients without ARDS (p < 0.002). The mean SAPS II scores on admission for the ARDS and non-ARDS patients (46 ± 18) did not differ.

Diagnosis of Pneumonia and Microbiology of Infection

Of the 73 episodes of VAP that occurred in the 49 of 134 patients with ARDS, 24 were diagnosed through BAL, seven by PSB culture, 25 by both techniques, and 17 by PTC culture. A total of 89 microorganisms were cultured at significant concentrations. The organisms recovered are listed in Table 4. The organism most frequently isolated was Pseudomonas aeruginosa (31%), followed by Staphylococcus aureus (21%), Enterobacteriaceae (14%), and Acinetobacter baumannii (10%). Nonfermenting gram-negative rods accounted for 47% (42 of 89) of the pathogens, with no significant difference between the numbers of these organisms found in first episodes (25 of 58) and further episodes of pneumonia (17 of 31). As indicated in Table 4, 84% of S. aureus were resistant to methicillin, and 18 of 73 (24%) of the episodes of pneumonia were polymicrobial. Nonfermenting gram-negative bacilli were significantly more frequent in patients with ARDS (42 of 89 pathogens in ARDS patients and 75 of 227 pathogens in non-ARDS patients, p = 0.02). In contrast, methicillin-susceptible S. aureus was significantly more frequent in non-ARDS patients (p = 0.008). There were no significant differences for other pathogens (Table 4).

Mortality and Duration of Mechanical Ventilation

Over half (78 of 134, 58%) of the ARDS patients died in the ICU (Table 2), as compared with 292 of 744 (39%) of the patients without ARDS who were on mechanical ventilation during the same study period (p < 0.001). However, there was no difference in the mortality rates of ARDS patients with and without VAP. Patients ventilated for reasons other than ARDS during the same period had identical mortality rates, whether they had pneumonia (67 of 162 patients) or not (225 of 582 patients; p = 0.5). The mean duration of mechanical ventilation for ARDS patients who did not develop pneumonia (11.3 ± 9.1 d) was the same as the mean duration of mechanical ventilation before the first episode of VAP for the patients who did develop pneumonia (11.7 ± 11.9 d; p = 0.8). However, the total time on mechanical ventilation was much longer for the patients with VAP (33 ± 21 d; p < 0.0001); the corresponding Kaplan-Meier estimates are shown in Figure 1. Similar findings were obtained when the analysis was restricted to the 56 patients who survived: the average total time on mechanical ventilation was 17 ± 12 d for surviving patients without pneumonia and 34 ± 15 d for surviving patients with VAP (Figure 2; p < 0.001).

View larger version (24K): [in this window] [in a new window]   Figure 1.   Duration of mechanical ventilation in the overall ARDS population. Patients with VAP remained on mechanical ventilation for significantly longer than those without VAP.

View larger version (23K): [in this window] [in a new window]   Figure 2.   Duration of mechanical ventilation for survivors from ARDS. Patients with VAP remained on mechanical ventilation for significantly longer than those without VAP.

Risk factors for VAP in ARDS patients

In addition to studying the parameters present at admission to the ICU and at the onset of ARDS, we studied the influence of other factors through the use of univariate analysis (Table 5). Sucralfate, which was administered to nearly 50% of the patients, was very significantly associated with pneumonia. The use of enteral nutrition was also associated with the occurrence of pneumonia, although less significantly. H₂-blockers were used in only 12 patients, and were associated with an apparent reduction in the incidence of pneumonia. However, the p values for the differences in the numbers of patients given H₂-blockers or enteral nutrition, respectively, could not be considered significant after Bonferroni's correction (Table 5). In contrast, the differences between patients with and without pneumonia in both the number of patients who received sucralfate and in the duration of sucralfate use remained highly significant. Patients who received sucralfate prophylactically had the same SAPS II values (48.3 ± 13.3 versus 43.8 ± 16, p = 0.07) and the same number of organ failures (2.49 ± 1.12 versus 2.69 ± 1.13, p = 0.3) upon entry into the protocol as those who did not receive sucralfate. The use of sucralfate, enteral nutrition, and H₂-blockers was entered into a logistic regression analysis. Both enteral nutrition and sucralfate were significantly associated (the latter with great significance) with pneumonia (Table 6). The duration of exposure to these potential risk factors was also considered. Univariate analysis (Table 5) showed that only the duration of sucralfate treatment was significantly associated with the occurrence of pneumonia. Logistic regression was performed, with enteral nutrition and H₂-blocker treatment forced into this model. The duration of sucralfate treatment remained very significantly associated with pneumonia (Table 7). Entering the values for SAPS II and for organ failure into the regression models did not alter the results (Tables 6 and 7).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Three conclusions can be drawn from this study. First, that VAP is frequent during ARDS. Second, that although pneumonia does not increase the mortality rate among mechanically ventilated patients, it markedly prolongs the time on mechanical ventilation. Third, that giving sucralfate to prevent gastroduodenal stress-related bleeding, a treatment usually considered to be safe, is a major predisposing factor for VAP. We believe this to be the first demonstration of such an effect of sucralfate.

Although four (7) recent studies evaluated the incidence of VAP during ARDS, none addressed the possible specific risk factors for such VAP, and all were conducted at single institutions. The diagnosis of pneumonia is often difficult to make when ARDS is present, since patients with ARDS have, by definition, diffuse abnormalities on chest radiographs. It may thus be very difficult to detect or interpret radiologic changes in such patients. Hence, it is crucial to obtain uncontaminated specimens of distal pulmonary secretions for quantitative culture. Although most of our patients were receiving antibiotics at the time of sampling, they were given the same regimen without any change for several days. Thus, any pneumonia that developed in these patients involved resistant pathogens (20). New antibiotics were not introduced before respiratory tract samples had been obtained. Thus, the incidence figure of 36.5% probably reflects the true incidence of VAP in ARDS patients. This figure is more than twofold that in the study done by Sutherland and colleagues (7). However, although these latter investigators used PSB and BAL to diagnose pneumonia, most of their patients underwent bronchospy at predetermined times during the course of ARDS, rather than when pneumonia was suspected. Thus, it is possible that some patients underwent bronchoscopy after the introduction of new antibiotics, increasing the probability of false-negative results (21). Chastre and coworkers (9) used similar diagnostic techniques and found a higher incidence (55%) of VAP among 56 patients with ARDS. However, many (45%) of the patients included in their study were on mechanical ventilation for postoperative respiratory failure, a situation in which the incidence of pneumonia may be higher (22). Delclaux and associates (8) prospectively evaluated lower respiratory tract infections in 30 patients with severe ARDS and found a much higher incidence of pneumonia (60%). They took repeated quantitative cultures of PTC specimens blindly, via the endotracheal tube, every 48 to 72 h after the onset of ARDS. However, blind PTC is probably less specific than other protected sampling techniques, because a significant number of bronchial samples are not distal and/or have a preferentially right-sided orientation (23). In addition, this technique was validated in only one study, done by the same investigators (18). In another study, Meduri and colleagues (10) used bilateral BAL in 94 ARDS patients and found a 43% rate of pneumonia, but did not compare the mortality and morbidity of patients with and without pulmonary infection.

The ICU mortality rate for the overall population of ARDS patients was high, in accord with previous findings (7- 9). This may have been because our patients had high SAPS II scores and a mean of two organ failures on admission, both of which are major determinants of mortality (24). Unsurprisingly, this mortality rate was significantly higher than that for mechanically ventilated patients without ARDS during the same period. However, we found that nosocomial pneumonia was not associated with an increased risk of mortality in ARDS patients. The lack of effect of VAP on mortality suggests that the antibiotic treatment of our patients was appropriate, but we did not specifically evaluate this issue. The increased mortality in mechanically ventilated patients resulting from nosocomial pneumonia remains a controversial issue, on which well-conducted studies have given conflicting results (1, 25). The extra mortality caused specifically by pneumonia in patients with severe ARDS is difficult to ascertain, since many patients die with multiorgan failure caused by a variety of etiologies (24).

Most investigators accept the link between the occurrence of nosocomial pneumonia and the increase in duration of mechanical ventilation. Papazian and coworkers (25) observed that surviving patients with pneumonia were on mechanical ventilation for 9 d longer than control patients without pneumonia. Our patients who developed pulmonary infection had exactly the same clinical and laboratory characteristics on admission as those who remained free of pneumonia. Thus, our study clearly indicates that the occurence of VAP was responsible for a nearly threefold increase in the duration of mechanical ventilation. It can be argued that those patients ventilated for the longest time have a higher risk of developing pulmonary infection. However, we showed that pneumonia was the cause of the increased time on ventilation. The time on mechanical ventilation until the first episode of pneumonia for patients who developed pneumonia was identitical to the time on ventilation until extubation or death for patients without this complication. The same was true for patients who survived their ARDS.

Nonfermenting gram-negative rods, especially P. aeruginosa, A. baumannii, and Stenotrophomonas maltophilia accounted for 47% of the responsible pathogens in cases of VAP, and 84% of the S. aureus strains isolated were resistant to methicillin. As did others, we found a high percentage (24%) of polymicrobial pneumonia (9). Our patients had two risk factors for VAP with resistant pathogens (20): First, 90% were receiving antibiotics (broad spectrum for most) at the onset of ARDS because of the severity of their underlying conditions (26). Second, VAP occurred late during the course of mechanical ventilation (20). The antibiotics given to nearly all of our patients explain the low incidence of early-onset pneumonia, which is usually caused by more susceptible pathogens (27). Additionally, the long duration of mechanical ventilation probably explained why more than 50% of our patients had several episodes of pneumonia.

Ours is the only study of VAP during ARDS that evaluated risk factors for such pneumonia. Surprisingly, we found that the use of sucralfate (and the duration of its use) was very significantly associated with the occurrence of pneumonia. To our knowledge, this is the first demonstration of such an association. Enteral nutrition and sucralfate (and its duration of use) were associated with pulmonary infection by univariate analysis in our study. Using a conservative approach, we first applied a Bonferroni's correction to account for the multiplicity of comparisons. Only the use (and duration) of sucralfate remained significantly associated with pneumonia (Table 5). The effect of sucralfate was yet more clear when all of the variables we examined were entered into a logistic regression analysis. These results were not affected when the clinical characteristics of patients receiving and not receiving sucralfate were taken into account. There was also a minor possible association between enteral nutrition and pneumonia, as already reported (28).

Several studies of the role of anti-stress-ulcer prophylaxis have found that agents that increase gastric pH are associated with an increased risk of VAP by comparison with sucralfate (11, 12, 29). In contrast, a recent study challenged this conclusion by showing the same rate of VAP in patients receiving ranitidine or sucralfate (30). It is noteworthy that the effect of sucralfate itself on the incidence of pneumonia remains unclear. However, a recent metaanalysis showed a trend toward a higher incidence of pneumonia among patients receiving sucralfate than among those given no prophylaxis (31). We have found only one large study that assessed the effect of sucralfate as compared with placebo on the incidence of pneumonia (32). Although the results fell short of significance, sucralfate tended to increase the rate of VAP (12% versus 6% in controls; relative risk: 2.0; CI: 0.79 to 5.01). A very large epidemiologic survey showed that all types of stress-ulcer prophylaxis (including sucralfate) increased the risk of pneumonia (33). Thus, our findings of an increased incidence of pneumonia with sucralfate in a selected population of ARDS patients are not totally unexpected. However, this potential risk factor was not assessed in the non-ARDS population in our study. The mechanism by which sucralfate may increase the rate of VAP in ARDS patients is unknown. Several publications have shown that sucralfate contributes to the constitution of pharmacobezoars (34), and esophageal and intestinal obstruction have been reported after sucralfate administration (35). This may favor bacterial proliferation, and hence increase the risk of pneumonia. Although it is difficult to draw definitive conclusions from a single study, our findings should prompt a reevaluation of sucralfate use.

The data obtained from the present large, prospective, multicenter study suggest that the incidence of VAP diagnosed by invasive techniques is higher than previously described (7), provided that a high index of suspicion is maintained throughout the course of mechanical ventilation. Because ARDS patients need prolonged mechanical ventilation and are usually treated with broad-spectrum antibiotics, pneumonia is often caused by difficult-to-treat pathogens. Nosocomial pneumonia does not seem to affect survival in this population at high risk for mortality. However, VAP considerably increases the time on mechanical ventilation. Moreover, and in contrast to the results obtained in previous studies, we found that patients given sucralfate may be at greater risk of developing pulmonary infection.