INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Histologic pneumonia is a frequent postmortem finding in nonsurvivors of acute respiratory distress syndrome (ARDS) (1), and clinical manifestations suggesting ventilator-associated pneumonia (VAP) invariably develop during the course of mechanical ventilation (2, 5). In patients with ARDS, a decision regarding the site of bronchoscopic sampling (BS) is made difficult by the presence of diffuse densities on chest radiograph. We have previously reported that bilateral bronchoalveolar lavage (BAL) sampling improves the recovery of opportunistic and nonopportunistic pathogens in immuno-compromised patients with diffuse radiographic densities (6, 7). In a pilot study of ARDS patients with suspected VAP, we found bilateral BAL to be positive on one side only in 7 of 15 (47%) episodes of VAP (8). Based upon this finding we implemented bilateral BAL sampling as our standard practice for evaluating suspected VAP in this patient population in our intensive care units.

To clarify clinical issues related to presentation, diagnosis, and outcome of pneumonia in ARDS, we designed a prospective data entry protocol and created a centralized BAL laboratory for rapid and consistent processing and analysis of BAL specimens. In ARDS patients subjected to diagnostic bronchoscopy, we followed recently developed guidelines (9) and, when tolerated, performed bilateral BAL sampling. BAL was processed for quantitative cultures and total cell count, and was subjected to microscopic analysis for cell differential, presence of intracellular organisms (ICO), and Gram stain. Microscopic analyses were performed by a single experienced technician. Our primary objective was to evaluate the diagnostic yield of bilateral BAL sampling in ARDS patients with suspected VAP, and to compare BAL microscopic and microbiologic results from contralateral sites. The secondary objective was to compare clinical characteristics and BAL results in patients with significant unilateral versus bilateral growth.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients

This study was conducted between January 1991 and March 1995 in the medical intensive care units (ICU) of the Regional Medical Center and The University of Tennessee Medical Center, Memphis, Tennessee. A centralized BAL laboratory with a full-time, dedicated laboratory technician (T.S.) was created for rapid and consistent processing of BAL specimens, other than microbiologic cultures. During the week, BAL specimens were processed in the centralized BAL laboratory within minutes of collection. During weekends, the specimens were processed by the hospital laboratory. A data collection form was created for prospective data entry into a FileMaker PRO (Claris, Santa Clara, CA) program designed for investigating clinical and laboratory questions related to pneumonia in ARDS.

We prospectively studied 111 consecutive patients who met diagnostic criteria for ARDS (10) and underwent diagnostic bronchoscopy for suspected VAP. Criteria for ARDS included the following: (1) respiratory failure requiring endotracheal intubation with acute onset of bilateral pulmonary densities on chest radiograph in the context of appropriate predisposing injury or illness with no evidence of left ventricular failure, and (2) a ratio of partial pressure of arterial oxygen to partial pressure of alveolar oxygen (Pa_O2:Fi_O2) of  0.2 (10). Clinical criteria for VAP in patients on mechanical ventilation (MV) for > 48 h included fever ( 38.3° C) and one or more of the following: new or progressive densities on chest radiograph; leukocytosis (> 10,000 cells/mm); or a macroscopically purulent tracheal aspirate. Antibiotic treatment (AB) at the time of bronchoscopy was classified following consensus guidelines (11) as no antibiotic use for greater than 48 h or less than 48 h and antibiotic use for greater than 48 h or less than 48 h. The type of antimicrobial agent at time of bronchoscopy and the indication for treatment were recorded. Antibiotic therapy following bronchoscopy was recorded.

Bronchoscopy and Specimen Processing

Premedication, monitoring, and exclusion criteria in patients at high risk for complications during bronchoscopy followed consensus guidelines (9). The fiberoptic bronchoscope (FOB) was inserted through the endotracheal tube ( 8 mm) via a sterile connector. Suction and injection of lidocaine through the bronchoscopic channel were avoided. Bronchoscopic samples were obtained from areas of new or progressive densities or from subsegmental bronchi with visible purulent secretions. The right middle lobe and lingula were selected if no roentgenographic changes or visible purulent secretions were observed. BAL was performed, following consensus guidelines (9), by introducing 140 ml of sterile nonbacteriostatic saline into a lung subsegment and aspirating back after each one of the aliquots. The return of the first 20-ml aliquot (bronchial fraction) was discarded. Unless abnormal findings were predominantly observed in the left lung, the right lung was sampled first. After the first BAL, the FOB was removed, the external surface of the FOB was cleansed with 70% isopropyl alcohol, and the suction lumen was flushed with 100 ml of sterile saline to minimize cross contamination between right and left BAL samples. Right and left BAL samples were collected and analyzed separately. Processing of BAL for quantitative cultures and microscopic analysis was previously described (12). BAL Giemsa stain for cell differential and percentage of ICO was prospectively reviewed by a single blinded observer (T.S.).

Data Analysis

BAL cultures were classified according to significant and insignificant growth. The diagnostic threshold for significant growth was  10⁴ colony-forming units/ml (cfu/ml). The BAL Gram stain was defined as negative if no microorganism was identified, true-negative if the ipsilateral BAL specimen had insignificant growth, or false-negative if the ipsilateral BAL specimen had significant growth. The Gram stain was defined as positive if microorganisms were identified, true-positive if they agreed with the morphology of all microorganisms grown in significant concentration in the ipsilateral BAL specimen, and false-positive if the BAL specimen had insignificant growth. The BAL Giemsa stain was defined as positive if  2% of the total inflammatory cells recovered contained ICO. Criteria for true and false Giemsa stain results were identical to those described for Gram stain.

Among bronchoscopies with significant bilateral bacterial growth on BAL, agreement between right and left BAL quantitative cultures was defined as complete (similar type of microorganisms with significant growth on both sides), partial (some but not all microorganisms grew in significant concentration in both right and left BAL), or absent (different microorganisms grew in significant concentration in both right and left BAL).

Statistical Analysis

Continuous and ordinal data such as microorganism load, percentage of neutrophils, and percentage of ICO were analyzed using the two-tailed t test for independent samples. A p value greater than 0.05 was considered nonsignificant (NS). For categorical data such as direct cause of ARDS, 10-d mortality and bacterial growth  10⁶ cfu/ml were analyzed using Mantel-Haenszel ² test and Fisher exact test when expected cell frequency was < 5.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Ninety-four of the 111 patients underwent bronchoscopy with bilateral bronchoscopic sampling. A total of 172 bronchoscopies with bilateral sampling were performed for a total of 344 BAL. Fifty-three patients were male and 41 were female. The mean (± SD) age was 45 ± 14 yr. Mean (± SD) Acute Physiology and Chronic Health Evaluation II (APACHE II) score on admission was 26 ± 7. ARDS was caused by a direct lung injury in 55 patients. We observed 14 complications (8%) attributable to bronchoscopy which included: transient oxygen desaturation (6), transient reduction in heart rate (3), transient reduction in blood pressure (1), minimal bronchial bleeding (1), transient increase in blood pressure (1), pneumothorax (1), cardiopulmonary arrest requiring cardiopulmonary resuscitation for 5 min (and patient expired with subsequent death 12 h after bronchoscopy due to progressive sepsis and multiple-organ dysfunction syndrome [MODS]) (1).

Forty (43%) patients had one or more episodes of pneumonia. Among 55 positive bronchoscopies, 33 (60%) had significant growth in only one side, 18 were right BAL, and 15 were left BAL. The false-negative rate for unilateral sampling varied from 32% (left BAL) to 26% (right BAL). Seven (13%) episodes of pneumonia were polymicrobial. Among 65 microorganisms recovered in significant concentration, Pseudomonas aeruginosa occurred in 43% and Staphylococcus aureus in 15% (Table 1). Overall, Gram stain had a sensitivity of 54% and a specificity of 87%; and Giemsa stain (> 2% ICO) had a sensitivity of 46% and a specificity of 93%. The mean percentage (± SD) of neutrophils in the 77 BALs with significant growth was 67% ± 29% and in the 267 BALs without significant growth was 60% ± 28%, respectively (p = 0.045). Among the 33 episodes with significant growth on only one side, the mean percentage (± SD) of neutrophils in BALs with significant growth was 54% ± 30% and in BALs without significant growth was 50% ± 28%, respectively (p = 0.51).

Bilateral Pneumonia

Table 2 shows the clinical characteristics in patients with significant bacterial growth on both sides compared with one side only during the first episode of VAP. No significant differences were found for any of the analyzed variables. Table 3 compares BAL results in episodes with significant bacterial growth on both sides versus one side only. Episodes of VAP with bilateral growth were more likely to be polymicrobial, have a bacterial growth  10⁵ cfu/ml, and have a higher percentage of neutrophils and ICO within inflammatory cells. In patients with significant bilateral bacterial growth, agreement between right and left BAL quantitative cultures was complete in 82% and partial in 18%. As shown in Table 1, bacteria identified most frequently in episodes with unilateral significant growth included P. aeruginosa 18, Acinetobacter calcoaceticus 3, Enterobacter cloacae 1, and S. aureus 4. 

Antibiotic Therapy before and after Bronchoscopy

Forty-five bronchoscopies were performed in patients no longer taking antibiotics; 25 patients were off for > 48 h. A total of 127 bronchoscopies were done in patients on antibiotics for 7.6 ± 5.2 d. Among the 107 bronchoscopies performed in patients receiving antibiotics for > 48 h, 96 were done in patients receiving treatment for an identified infection and 11 were in patients receiving empiric antibiotic therapy. Among the 20 bronchoscopies in patients on antibiotics for < 48 h, 15 received treatment for an identified extrapulmonary infection. The overall effects of antibiotics on Gram stain and Giemsa stain sensitivity and specificity are shown in Table 4. In patients on antibiotics > 48 h, the sensitivity of Giemsa stain was lower in episodes with significant growth in one side only compared with those with growth in both sides (22% versus 56%; p = 0.034), while the yield of Gram stain was similar (53% versus 52%). The percentages of complete agreement for Gram stain and Giemsa stain were similar in patients off antibiotics for > 48 h and on antibiotics for  48 h.

In 50 of the 55 (91%) episodes of VAP, antibiotic treatment was modified after the results of BAL microscopic analysis and bacterial cultures became available (Table 5). In 37 episodes a new antibiotic was added and in 15 episodes all antibiotics were changed. Of the 117 episodes with negative cultures, 30 were in patients off antibiotics at the time of bronchoscopy. In this group, six continued off antibiotics and 24 received subsequent antibiotic treatment (six for an identified and 18 for a presumed extrapulmonary infection). Of the 87 episodes in patients on antibiotics (70 for > 48 h) at the time of bronchoscopy and negative BAL cultures, four had antibiotics discontinued, 21 had new antibiotics added (10 for an identified and 11 for a presumed extrapulmonary infection), nine had antibiotics changed (three for an identified and five for a presumed extrapulmonary infection), and 53 had continuation of the prior treatment (39 for an identified and 14 for a presumed extrapulmonary infection).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The main findings of this prospective study of patients with ARDS and suspected VAP subjected to bilateral BAL are: (1) most episodes (68%) had bilateral insignificant bacterial growth, (2) the recovery of bacteria in significant concentration ( 10⁴ cfu/ml) was frequently unilateral (60%), (3) episodes with bilateral significant growth were more likely to be polymicrobial, to have a higher bacterial colony count, and a more intense local inflammatory response. Clinical manifestations suggesting pneumonia are a frequent occurrence during ARDS, and postmortem studies have consistently identified bronchopneumonia by strict histologic criteria in 48 to 73% of ARDS nonsurvivors (1). When compared with histological findings, however, clinical predictors of pneumonia are unreliable (2, 13). The severe and protracted pulmonary inflammatory response that characterizes ARDS (14) invariably causes fever and leukocytosis in the absence of infection (13). During ARDS, fever and leukocytosis develop with an equal frequency and degree in patients with and without infection (5). Studies using open lung biopsies in patients with unresolving ARDS have shown that (a) pneumonia is an infrequent (20%) histologic finding (15), (b) clinical and radiographic criteria cannot differentiate VAP from the progression of fibroproliferation (13, 16), (c) colonization of the upper and lower airways with gram-negative bacteria (17) makes routine culture of the endotracheal aspirate nonspecific (56% false-positive rate) (15), and (d) lack of significant bacterial growth in BAL obtained prior to lung biopsy correlates with histologic absence of pneumonia (18). Bronchoscopic sampling of the distal airways is advocated to improve the diagnosis of VAP and has been used to guide antimicrobial therapy (12). In patients with ARDS, a decision regarding the site of bronchoscopic sampling is made difficult by the presence of diffuse densities on chest radiograph. Bedside chest radiographs or computer tomography of the chest are not helpful in predicting who has pneumonia (19, 20). Improving our ability to diagnose pneumonia in patients with ARDS is important for both managing patients and conducting reliable clinical investigations. The following findings are reviewed in the context of the literature: distribution of bacteria in pneumonia and the role of bilateral sampling, relationship between bacterial load and local inflammatory response, value of microscopic analysis for early diagnosis of pneumonia, and effect of antibiotic treatment on the result of cultures.

Bilateral BAL theoretically doubles the sampling area of unilateral lavage. In the present study, 33 of 55 positive bronchoscopies had significant growth in one side only. These data suggest that sampling area may be an important critical variable in diagnosing VAP in patients with ARDS and agree with recent postmortem studies indicating that limited sampling with a single lung biopsy cannot exclude pneumonia (4, 21, 22).

Fabregas and coworkers (22), conducted an extensive postmortem examination in 25 patients on mechanical ventilation (five with ARDS) obtaining 375 biopsies for quantitative microbiology and histology. They found VAP to be a nonhomogeneous multifocal process showing different stages of evolution at the same time, and frequently involving both lungs (22). In a detailed description of postmortem lung tissue (one lung) involving 83 patients (most with ARDS), Rouby and coworkers (4) reported large areas of confluent pneumonia in 23 (28%) and small foci of bronchopneumonia disseminated within different segments in 14 (17%). Frequently, typical histologic aspects of bronchopneumonia were observed in only one or two slices obtained from the same lung segment, the implication being that inadequate sample size may cause a misdiagnosis. Histologic bronchopneumonia was identified in more than 80% of dependent lung segments and 50 to 70% of nondependent lung segments (4). In a primate model of ARDS, Johanson and coworkers (21) obtained postmortem lung tissue histology and quantitative bacterial cultures of multiple lobes and found a wide variation in the bacterial burden. Animals without prior antibiotic treatment, a factor that may negatively affect culture results, had sterile lobes next to or contralateral to lobes with significant bacterial growth ( 10⁴ cfu/g) (21).

It is estimated that the alveolar surface area distal to the wedged bronchoscope is 100 times greater than that of the peripheral airway and that approximately 1 million alveoli (1% of the lung surface) are sampled. In an experimental model of ARDS, Johanson (21) found that in the absence of antibiotic therapy, BAL was the most sensitive methodology for recovery of bacteria, retrieving 74% of all species present in lung tissue (compared with 41% by protected specimen brushing [PSB]). Chastre and coworkers reported that in nonsurvivors of acute respiratory failure who had no recent change in antibiotic therapy and who were subjected to postmortem bronchoscopy and thoracotomy with multiple lung biopsies, all microorganisms present at a significant concentration ( 10⁴ cfu/ g) in tissue culture were also present in the BAL at a concentration of  10⁴ cfu/ml (23). In a recent study comparing postmortem lung tissue with bronchoscopic sampling, Kirtland and coworkers (24) have found sterile BAL to have a specificity of 96% and a positive predictive value of 91% for sterile lung parenchyma, a finding in agreement with a prior open lung biopsy report (18). In the above clinical studies, the specificity of unprotected BAL quantitative cultures varied from 63% (24) to 78% (23). This rate of false-positive results is in agreement with the literature and the possible mechanisms for BAL contamination were previously reviewed (25). Our large prospective evaluation of VAP in ARDS lacks histologic confirmation of pneumonia, and the limitations of BAL specificity should be taken into consideration when interpreting our results. Although the diagnostic threshold of 10⁴ cfu/ml used for defining a positive BAL quantitative culture may be too sensitive (12), this cutoff point was proposed by a consensus of investigators to avoid underdiagnosis of pneumonia (9). In performing bilateral BAL, we were concerned with possible contamination of the contralateral lung sampling. Right and left BAL had a similar rate of positive results, indicating that cleansing of the bronchoscope in between samplings may minimize the risk of contaminating the second BAL. No prior study has evaluated bilateral bronchoscopic sampling for diagnosing VAP during ARDS. In patients without ARDS and subjected to bilateral PSB, sampling of the radiographically involved segment and an uninvolved area in the contralateral lung showed a high level of agreement (87%) in one study (26) and lack of correlation (12.5%) in another (27).

BAL findings indicate that episodes with a higher bacterial burden had cytologic evidence of a more intense local inflammatory response (Table 3), and were more likely to be diffuse. Experimental and human studies have shown that a lung affected by ARDS is impaired in its ability to clear a bacterial challenge. Several intrinsic defects have been previously implicated, primarily those related to changes in the alveolar environment and the function of phagocytic cells (28). When VAP develops, clinical and postmortem studies have described a strong association between number of bacteria and severity of local inflammation (4, 21, 23). Polymorphonuclear cells recruited into the airspaces of patients with ARDS have shown evidence of impaired microbicidal activity (29, 30); this mechanism partly explains the lung's inability to clear bacteria in spite of intense local inflammation. Furthermore, polymorphonuclear leukocyte (PMN) clearing of bacteria is dose-dependent, and the efficiency of PMN bactericidal activity decreases with increasing bacterial load (31). Recovery of an identical microorganism in both lungs may be explained by either aspiration of a large bacterial load or by spread to the contralateral lung.

Because microscopic analyses of BAL can be available within a few hours of bronchoscopy, the recovery of a larger percentage of neutrophils and ICO can be used to indicate a more intense inflammatory reaction associated with a heavier bacterial burden. Our findings agree with previous studies indicating that prior antibiotic treatment received for an earlier infection unrelated to the suspected pneumonia does not affect the diagnostic yield of BAL bacterial cultures (32, 33).

Although bilateral BAL was well tolerated, a life-threatening complication developed, and this occurrence underscores the need for careful implementation of safety measures and monitoring to avoid unnecessary harm to the patient (9). This methodology is routine practice in our intensive care units and our policy is to terminate the procedure if oxygen desaturation or hemodynamic instability develops. In our study the incidence of VAP is higher than the one (15%) recently reported by Sutherland and colleagues (34). In that study, however, unilateral BAL results may have underappreciated the true incidence of pneumonia because bronchoscopy was performed at predetermined intervals and not in patients with suspected VAP; additionally, antibiotic treatment was not controlled prior to bronchoscopic sampling. Our findings are similar to recent reports from Chastre and coworkers (35), and Delclaux and coworkers (36).

In conclusion, this prospective study of patients with ARDS showed that in most (68%) episodes of suspected VAP, bilateral BAL quantitative culture had insignificant bacterial growth. When VAP was identified, the recovery of > 10⁴ cfu/ml of bacteria was frequently unilateral (60%). Our study confirms prior observations that pneumonia with a higher bacterial burden shows cytologic evidence of a more intense local inflammatory response, and is more likely to be diffuse. Bronchoscopic findings modified antibiotic therapy in 91% of patients with positive BAL cultures and avoided introduction of new antibiotics in 54% of patients with insignificant growth.