INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The best option for diagnosing pneumonia in patients receiving mechanical ventilation remains a subject of controversy. Following the recommendations of an international consensus conference (1), six studies recently evaluated the diagnostic values of various bronchoscopic and nonbronchoscopic techniques by using immediate postmortem pulmonary biopsy and culture as the gold standard for the final diagnosis (2). Regarding bronchoscopic techniques, neither definite nor unequivocal conclusions could be drawn from these studies. Indeed, in four of them (3, 5) it was concluded that quantitative cultures of protected specimen brush (PSB) and bronchoalveolar lavage (BAL) could hardly differentiate the histologic presence or absence of pneumonia. In contrast, Chastre and coworkers (4) showed that PSB and BAL could fairly well identify lung segments with active infection. These discrepancies may be due to differences in definition of pneumonia: defined by the presence of histologic lesions alone (3, 5, 7), by histologic lesions together with positive lung cultures (6), or by high bacterial counts in lung cultures (4). They may also result from difficulties encountered in definitely recognizing lesions of an ongoing pneumonia in patients with a previous episode of pneumonia or with associated noninfectious pulmonary lesions (2). Finally as stressed by Chastre (4), when using diagnostic techniques based on quantitative cultures in order to discriminate between presence or absence of pneumonia, it is crucial that no new antibiotics were introduced before PSB, BAL, and lung cultures are obtained.

To our knowledge only two studies (4, 7) evaluated the diagnostic accuracy of bronchoscopic and nonbronchoscopic techniques in establishing the bacteriologic etiology of ventilator-acquired pneumonia (VAP). Here again, conflicting results were reported which may be accounted for by differences in defining causative organisms and by differences in the antibiotic therapy in both series.

From previous experiments, we knew that piglets rapidly develop endogenous pneumonia as a complication of experimental bronchial stenosis (8). In this model of postobstructive pneumonia we showed that the parenchymal bacterial burden was unevenly distributed within the lungs and even within the lung segments and that no clear-cut threshold for quantitative cultures could discriminate the presence or absence of pneumonia. We also knew that piglets are prone to spontaneously develop VAP provided they are subjected to prolonged ventilatory support. With respect to evaluation of diagnostic techniques, this model of VAP offers several advantages compared with postmortem studies in humans. Evaluation can be performed in the absence of antibiotic therapy. Histologic examination is not confounded by previous episodes of pneumonia or by coexisting noninfectious lesions. Multiple samples can be obtained by means of bronchoscopy and lung biopsy. In this model of VAP we studied the distribution of total and single bacterial burden, the relationship between lung histology and bacteriology, and finally the influence of these features on the accuracy of diagnostic techniques.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Experimental Design

Random bred domestic Largewhite-Landrace piglets (22 ± 2 kg) were mechanically ventilated under general anesthesia for a duration of 4 d by means of a volume-controlled respirator (Monal D; Taema, Antony, France). General anesthesia, analgesia, and paralysis were produced by continuous intravenous infusion of midazolam (Hyphovel; Roche Laboratories, Nevilly-Sur-Seine, France; 0.3 mg/kg/h), fentanyl (Fentanyl; Janssen-Cilag, Boulogne, France; 5 µg/kg/h), and pancuronium bromide (Pavulon; Organon-Teknika, Saint Denis, France; 0.32 mg/kg/h). The antibiotic group (ATB group) received ceftriaxone (Rocephin; Roche Laboratories) 1 g intravenously twice daily from the start of mechanical ventilation as an attempt to prevent VAP. Another group (control group) underwent mechanical ventilation without antibiotics. This study was approved by the review board of the Department of Experimental Research of Lille University. All animals were treated in compliance with the guidelines of the Department of Experimental Research of Lille University and with the Guide for the Care and Use of Laboratory Animals (NIH Publication No. 93-23, revised 1985).

Bacteriological Samplings

On Day 1, BAL was performed in the lingular bronchus and served as control. After careful tracheobronchial suctioning a fiberoptic bronchoscope was introduced into the endotracheal tube through an adaptor designed to minimize air leak (Model 514900; Rüsch AG, Kernen, Germany) and advanced under direct vision into the lingular bronchus without suction, nor use of topical anesthetics. A first aliquot of 20 ml of sterile saline was instilled, gently aspirated ("bronchial fraction") and discarded. Three additional aliquots of 20 ml were instilled, reaspirated, pooled, and processed for further analysis (quantification of alveolar cell count and differential and bacteriologic studies). On Day 4, endotracheal aspirates were obtained by careful endotracheal suctioning using a sputum suction trap (Model 534-16; Vygon, Ecouen, France) for direct examination and quantitative cultures (QEA). Fiberoptic bronchoscopy was then performed for PSB and BAL. PSB and BAL specimens were collected from the right middle lobe bronchus and the apical bronchus of the right lower lobe.

Collection of Lung Tissue Specimens

After the collection of bronchoscopic specimens the animals were killed. Six superficial lung tissue specimens (approximately 1 cm³ each) were excised from the most dependent segments (lingula, middle lobe, anterior segments of the left and right lower lobes) and from the most "nondependent" segments (apical segments of the left and right lower lobes) while mechanical ventilation was maintained. Sampling was always performed in areas showing gross abnormalities, when present. Each specimen was cut in two parts in vis-à-vis (one for quantitative cultures and one for histologic study) in order to properly compare histologic and bacteriologic findings.

Bacteriologic Processing of Specimens

QEA, PSB, BAL, and lung tissue specimens were processed for microscopic examination and bacterial cultures as previously described, according to recommended laboratory methods (5, 8). For lung tissue specimens, counts of each identified bacterial species were expressed in colony-forming units per gram (cfu/g) of tissue. In addition, for each specimen the total number of bacteria was calculated by adding the absolute number of bacteria cultured from the specimen and the result was expressed in cfu/g of tissue.

Pathologic Study

Specimens were processed according to standard methods. Evaluations were made by two observers, independently, without knowledge of animal and bacteriologic data. The lesions were graded as previously described (2, 5, 8) into six categories: no pneumonia (NP), purulent mucous plugging (PMP), bronchiolitis (B), pneumonia (P), confluent pneumonia (CP), and abscessed pneumonia (AP). Classification of each specimen was based on the worst category observed. The diagnosis of pneumonia included only the P, CP, and AP categories.

Data Analysis

Data are presented as mean ± SD, except as otherwise specified. The Fisher exact test was used to compare categorical variables. For continuous variables the Mann-Whitney test for unpaired series was used. Comparisons of measured parameters within each group were assessed by two-way repeated measures analysis of variance (ANOVA) or by the Wilcoxon test for paired series depending on the size of the sample. Correlation was assessed using the Spearman rank test. A value of p < 0.05 was considered to indicate statistical significance. Sensitivity and specificity of each diagnostic technique were calculated according to standard formulas.

Recommended diagnostic thresholds (10³ cfu/ml of Ringer's and 10⁴ cfu/ml of retrieved fluid) were used for quantitative cultures of PSB and BAL, respectively. For QEA, two diagnostic thresholds were studied (10⁵ and 10⁶ cfu/ml of respiratory secretions). For direct examination, PSB and EA specimens were considered as positive if screening for microorganisms at high magnification (100-fold) was positive on the May-Grünwald-Giemsa (MGG) and Gram stained slide (11). BAL specimens were considered as positive if the percentage of cells containing intracellular organisms (ICO) was  1 % (5).

The ability of diagnostic techniques to differentiate the histologic presence or absence of pneumonia was assessed as follows: operative values of each diagnostic technique were first calculated by taking as reference diagnosis the histology of the lung area where the bronchoscopic sample was obtained (histologic presence or absence of pneumonia in the segment). For instance, in an animal without histologic evidence of pneumonia in the right middle lobe, a PSB specimen collected in the right middle lobe bronchus and growing < 10³ cfu/ml of Ringer's was classified as "true negative" whatever the findings (presence or absence of pneumonia) in other areas of the lung. Calculations were also made by taking as reference diagnosis the histology of the lungs (histologic presence or absence of pneumonia in the animal). For instance, in an animal showing histologic evidence of pneumonia in at least one segment, a PSB specimen collected in the right middle lobe bronchus and growing  10³ cfu/ml of Ringer's was classified as "true positive" if pneumonia was present in the right middle lobe but also if pneumonia was not present in that lobe.

In control animals with histologically proven pneumonia, the ability of diagnostic techniques to identify the causative organisms (bacteriologic diagnosis) was assessed as follows: an organism was considered as causative of the pneumonia if it was cultured at a concentration  10⁴ cfu/g of tissue. A specimen that missed one of the causative organisms (causative organisms absent or present only at a concentration below the recommended diagnostic threshold) was considered as a "false negative."

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Population

Twenty-seven animals completed the 4-d study protocol (18 control and 9 ceftriaxone-treated animals). A total of 144 pulmonary biopsy specimens were available for paired histologic and bacteriogical analysis. For technical reasons, several planned pulmonary biopsies, PSB and BAL specimens were not obtained. As a whole, in control animals, 34 PSB and BAL specimens (collected in 18 dependent and 17 nondependent lung segments) were available. In the ATB group, 16 PSB and 17 BAL specimens (collected in nine dependent and eight nondependent lung segments) were available.

Histologic and Cytologic Findings

All animals but one in the control group (17 of 18) and 4 of 9 animals in the ATB group developed histologically proven pneumonia (p < 0.05). Pneumonia was most often bilateral and clearly predominated in dependent lung segments. Histologic signs of pneumonia were respectively present in 60% and 17% of the studied segments in control and ceftriaxone-treated animals (p < 0.001).

Compared with BAL performed on Day 1, BAL performed on Day 4 in segments with histologically proven pneumonia showed a dramatic increase in cellularity (655 ± 370 cells per mm³ versus 5,305 ± 4,400 cells per mm³; p < 0.001) and neutrophil counts (21 ± 22 cells per mm³ versus 4,260 ± 3,970 cells per mm³; p < 0.001).

Bacteriologic Findings

In control animals, mean bacterial count was 4.5 × 10⁵ ± 2 × 10⁵ cfu/g tissue in pulmonary biopsy specimens with histologically proven pneumonia and 7.9 × 10⁴ ± 4.9 × 10⁴ cfu/g tissue in lung specimens without pneumonia (p < 0.01). In ceftriaxone-treated animals, mean bacterial count was 7.8 × 10² ± 7.1 × 10² cfu/g tissue in pulmonary biopsy specimens with pneumonia and 7.5 × 10² ± 2 × 10² cfu/g tissue in specimens without pneumonia (nonsignificant, NS).

In 17 of 21 cases of pneumonia, lung cultures yielded a mixed flora with an average of 2.05 and 1.12 organisms per biopsy in the control and ATB group, respectively. In only four cases the pneumonia was monomicrobial (one in the control group and three in the ATB group). Pasteurella multocida and Streptococcus suis were the predominating organisms.

Relationship between Bacterial Burden and Histologic Lesions

Although, at least in the control group, the total number of bacteria grown in pulmonary biopsy specimens correlated with the corresponding histologic grades, this relationship was not straightforward (Figure 1). No clear-cut bacteriologic cutoff could be defined to differentiate the histologic presence or absence of pneumonia.

View larger version (22K): [in this window] [in a new window]   Figure 1.   Relationship between the total number of bacteria grown in pulmonary biopsy specimens and the corresponding histologic grades.

Distribution of Bacterial Burden among the Studied Lung Areas

As shown in Figure 2, the total number of bacteria varied markedly from one area to another. This was true when comparing matched left and right segments (Figure 2A), but also when comparing immediately adjacent lung segments such as the lingula and the anterior segment of the left lower lobe or the right middle lobe and the anterior segment of the right lower lobe (Figure 2B). This marked variation was also observed when considering the distribution of bacterial concentrations of single bacterial species (Figure 3).

View larger version (31K): [in this window] [in a new window]   Figure 2.   Comparisons of the bacterial burden present in matched (A) or immediately adjacent (B) areas of the lungs. Each symbol represents a single lung segment. (A) Scatterplot of the relationship between the quantitative culture results (log10 cfu/g of tissue) for all bacterial species recovered from a given segment and its corresponding contralateral segment. (B) Scatterplot of the relationship between the quantitative culture results (log10 cfu/g of tissue) for all bacterial species recovered from immediately adjacent lung segments.

View larger version (32K): [in this window] [in a new window]   Figure 3.   Comparisons of the concentrations of single bacterial species in matched (A) or immediately adjacent (B) areas of the lungs. Each symbol represents a single bacterial species. (A) Scatterplot of the relationship between the quantitative culture results (log10 cfu/g of tissue) for each bacterial species recovered from a given segment and its corresponding contralateral segment. (B) Scatterplot of the relationship between the quantitative culture results (log10 cfu/g of tissue) for each bacterial species recovered from immediately adjacent lung segments.

Diagnosis of Pneumonia by Endotracheal Aspirates (EA), PSB, and BAL

Sensitivities and specificities of EA, PSB, and BAL in recognizing histologic presence or absence of pneumonia are reported in Table 1. Although the difference never reached statistical significance, the sensitivities of bronchoscopic techniques were always lower when calculated by taking whole lung's histology as reference diagnosis than when calculated by taking the segment's histology as reference diagnosis. This was simply because pneumonia was not homogeneously distributed through the lungs so that in a certain number of animals with pneumonia the segments studied by PSB or BAL were free of pneumonia. The low specificity (high false-positive rate) of quantitative cultures of PSB could potentially be explained in three ways: among nine false-positive PSB specimens (growing  10³ cfu/ml from a segment without pneumonia) the culture of that segment grew  10³ cfu/g tissue in six cases and in two cases the anterior segment of the right lower lobe (which is adjacent to the right middle lobe) grew  10³ cfu/g tissue. No potential explanation was found to explain the false positivity in one case. The same explanations held true for the nine false-positive BAL specimens: the culture of the segment grew  10³ cfu/g tissue in eight cases and in one case the adjacent segment grew  10³ cfu/g tissue. False-negatives (PSB and BAL specimens growing respectively < 10³ cfu/ml and < 10⁴ cfu/ml from a segment with pneumonia) could be explained in three ways: in ceftriaxone-treated animals the culture of the segment (despite histologic presence of pneumonia) was sterile in all false negatives. In control animals, although in all these cases the segments grew  10³ cfu/g tissue, three PSB and one BAL specimens were sterile. Four PSB and one BAL specimens had quantitative culture results below the recommended diagnostic thresholds.

Identification of the Causative Organisms by EA, PSB, and BAL

In both dependent and nondependent lung segments the total number of bacteria obtained from PSB and BAL specimens significantly correlated with the total number of bacteria in the corresponding lung segment (Figure 4). Despite this, except for QEA, the ability of diagnostic techniques to identify the causative organisms (i.e., to establish the bacteriologic diagnosis of pneumonia) appeared particularly low in control animals.

View larger version (32K): [in this window] [in a new window]   Figure 4.   Relationship between the total number of bacteria recovered from biopsy specimens of the middle lobe (left) and the apical segment of the right lower lobe (right) and the corresponding PSB (top) and BAL (bottom) samples.

For PSB, when the lung's bacteriology was taken as reference 11 of 31 specimens (35%) correctly identified all the causative microorganisms (cultured at a concentration  10⁴ cfu/g tissue). Among the 20 false-negative specimens, 14 completely missed the causative organisms and six grew the causative organisms but at a concentration below the 10³ cfu/ml diagnostic threshold. When the segment's bacteriology was taken as reference 12 of 24 PSB specimens (50%) correctly identified all the causative microorganisms in the corresponding segment. Among the 12 false-negative specimens, seven completely missed the causative organisms and five grew the causative organisms but at a concentration below the 10³ cfu/ ml diagnostic threshold.

For BAL, when the lung's bacteriology was taken as reference 16 of 32 specimens (50%) correctly identified all the causative microorganisms. Among the 16 false-negative specimens, eight completely missed the causative organisms and eight grew the causative organisms but at a concentration below the 10⁴ cfu/ml diagnostic threshold. When the segment's bacteriology was taken as reference 21 of 28 BAL specimens (75%) correctly identified all the causative microorganisms in the corresponding segment. Among the eight false-negative specimens, five completely missed the causative organisms and three grew the causative organisms but at a concentration below the 10⁴ cfu/ml diagnostic threshold.

For QEA, only the lung's bacteriology could be taken as reference. When the 10⁵ cfu/ml diagnostic threshold was considered, 16 of 17 QEA specimens (94%) correctly identified all the causative microorganisms. The only false-negative specimen completely missed the causative organisms. When the 10⁶ cfu/ml diagnostic threshold was considered, 13 of 17 QEA specimens (76%) correctly identified all the causative microorganisms. Among the four false-negative specimens, one completely missed the causative organisms and three grew the causative organisms but at a concentration below the 10⁶ cfu/ ml diagnostic threshold.

These figures regarding the false-negative rates for each diagnostic technique only slightly decreased when changing the definition of causative organisms to "organism(s) cultured at a concentration  10³ cfu/g of tissue" (instead of  10⁴ cfu/g).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Our results indicate the following: (1) Although the local bacterial burden tends to correlate with the severity of histologic lesions, no definite bacteriologic cutoff can differentiate the histologic presence or absence of pneumonia. (2) As is the case for histologic lesions, the total bacterial burden was unevenly distributed through the lung. (3) This heterogeneity in bacterial distribution also holds true for single bacterial species. (4) Using discriminative values ( 10³ cfu/ml,  10⁴ cfu/ ml, and  10⁵ cfu/ml) to define positive PSB, BAL, and QEA cultures respectively, these techniques identified the histologic presence of pneumonia with a sensitivity of 69, 78, and 100%, respectively. (5) In contrast, the specificity of these techniques in recognizing VAP is less than 50%. (6) Finally, when using usual diagnostic thresholds, less than 50% of PSB and BAL specimens correctly identified the causative organisms whereas 94% of EA specimens correctly established the microbiologic diagnosis of pneumonia.

Some of the difficulties in evaluating diagnostic procedures in human VAP can be overcome in animal models. Indeed, in contrast to humans, antibiotics can be withheld until death in animals with suspected pneumonia, multiple diagnostic procedures can be applied and thus compared, and lung tissue can easily be obtained for histologic and bacteriologic examination in the absence of previous antibiotic therapy and pre- or coexisting pulmonary illnesses. Before PSB had been evaluated in mechanically ventilated patients, Higuchi studied a primate model of oleic acid acute diffuse lung injury with naturally occurring pneumonia (12) and Moser studied a canine model of Streptococcus pneumoniae pneumonia obtained by exogenous bacterial challenge (13), showing that PSB obtained uncontaminated specimens from the distal airways in intubated animals. Later, the former authors (14) showed in the primate model that, when compared with endotracheal aspirates and PSB, BAL "provided the best reflection of the lung's bacterial burden, both quantitatively and qualitatively in the setting of prolonged intubation and ventilation." Like the baboons described by the group of Johanson (14) our piglets subjected to prolonged ventilation spontaneously developed an endogenously acquired pneumonia which resembles human VAP. We therefore studied how the relationship between histology and bacteriology could influence the accuracy of diagnostic techniques in this model of VAP.

Initially the PSB technique was developed to obtain uncontaminated respiratory tract secretions from the distal airways. As shown by the group from the University of South Alabama (15), in more than 90% of nonintubated patients with clinical evidence of bacterial pneumonia, PSB recovers high concentrations ( 10³ cfu/ml) of bacterial pathogens, provided patients are not on antibiotics by the time of the diagnostic procedure. Subsequently, PSB was proposed to identify pneumonia in mechanically ventilated patients (18). The use of PSB and BAL to recognize pneumonia in ventilated patients is based on two successive assumptions: first, quantitative cultures (QC) of bronchoscopic specimens provide a good reflection of the bacterial burden present in the lung parenchyma. Second, there is a clear-cut relationship between bacterial concentrations in lung parenchyma and the presence of pneumonia. Our results (Figure 4) agree with the findings of Chastre (4, 18) in humans and the findings of Johanson in primates (14) showing that the total number of bacteria obtained from PSB and BAL specimens and from the corresponding lung segment significantly correlate. Therefore, at least in the absence of antibiotics or of recent changes in antibiotics, the first assumption seems to be valid. On the contrary, the validity of the second assumption is questioned by several studies showing that there is no meaningful correlation between histologic features and QC of lung biopsies (2, 3, 6, 19). In contrast to these findings, three studies (4, 14, 18) classifying lesions of pneumonia into "mild, moderate or severe" (14) reported a significant correlation between the bacterial burden present in the lung and the severity of histologic lesions found in the same area and concluded that QC of pulmonary biopsies could recognize at least moderate and severe pneumonia in ventilated patients.

Several potentially confounding factors may have precluded definite conclusions regarding the relationship between severity of lung damage and local bacterial burden in VAP. First, except in one study by Chastre (4) many patients and many animals were receiving antibiotics or had recent changes in antibiotics. This may have lessened or negated lung QC. Second, most of the series in humans did not exclude the patients with a previous episode of pneumonia (2, 3, 6, 7). Thus, low bacterial counts together with histologic abnormalities may have represented a resolution phase of pneumonia. Third, except in the studies by Kirtland (7) and by Fabregas (19) specimens for bacterial culture and specimens for pathologic studies were not immediately adjacent. This complicates comparison between histology and bacteriology since VAP is a heterogeneous process, both in term of severity of histologic lesions and in term of anatomic distribution of the lesions. Fourth, as pointed out by Corley (20) the histologic definition of pneumonia also adds to the confusion. This is especially true when classifying the lesions into "mild, moderate or severe" (3, 4, 6, 7, 14, 18). "Scattered neutrophilic infiltrates localized to terminal bronchioles and some surrounding alveoli" used to define "mild pneumonia" (4, 14) can also be seen in presence of pre- or coexisting pulmonary lesions such as adult respiratory distress syndrome (ARDS), atelectasis, or congestive heart failure. The difficulty to attribute these signs to true bronchopneumonia may explain why in Chastre's study, among the 17 segments showing histologic evidence of "mild pneumonia," 11 grew less than 10⁴ cfu/g tissue or were sterile (4).

In this model of VAP we studied animals free of antibiotic therapy and animals receiving ceftriaxone prophylactically. All were free of any previous or concomitant lung disease. In order to add confidence to comparative analysis, lung samples for histologic and bacteriologic studies were taken vis-à-vis as proposed by Kirtland (7). We used the more clear-cut classification of histologic lesions proposed by Rouby (2). Finally, to compare our results to those obtained from human studies, we expressed culture results in cfu/g tissue instead of using the Bacterial Index as done by Johanson in the primate model (14). Our findings clearly show that, in the absence of the above listed confounding factors, quantitative lung cultures cannot reliably differentiate the presence or absence of pneumonia (Figure 1).

The last finding that questions the reliability of PSB and BAL in recognizing VAP is the large variation in distribution of the bacterial burden throughout the lung in a given subject. When comparing paired segments (anterior segment of the left upper lobe and anterior segment of the left lower lobe), Chastre and coworkers found that, although the total bacterial burdens significantly correlated, there were several cases where one segment yielded high bacterial counts whereas the matched segment yielded low bacterial counts or even was sterile (4). Similar observations were made in the present study when comparing the total bacterial burdens from matched left and right segments and even when comparing immediately adjacent segments.

Although many studies evaluated the capacity of PSB and BAL to recognize pneumonia in ventilated patients, few studies specifically addressed the issue of the microbiologic diagnosis of VAP with these techniques. This point is of crucial importance for clinical practice. For instance, in a patient with suspected VAP, a PSB yielding Escherichia coli 10⁴ cfu/ml and Staphylococcus aureus 10 cfu/ml would lead to the conclusion that the patient is most probably suffering pneumonia. However, the question regarding the causative organism (should E. coli alone or E. coli and S. aureus be considered for antibiotic therapy?) remains open. In the study by Kirtland (7) the sensitivity of PSB and BAL in identifying the microbial species cultured in lung tissue was 44% and 65%, respectively. In contrast, EA recognized the microbial species isolated from the lung with an 87% sensitivity. These figures were obtained while most of the patients were receiving antibiotics by the time of the diagnostic procedure and without considering bacterial concentrations in lung, PSB, BAL, and EA specimens. In the study by Chastre (4) where patients had discontinued antibiotics or had no recent changes in antibiotic therapy, of the 32 microbial species present in lung at a concentration 10⁴ cfu/g tissue, 29 (90%) were recovered by the brush at a concentration  10³ cfu/ml and all were also present in lavage fluid at concentrations  10⁴ cfu/ml. In the present study we used a similar definition for causative organisms (i.e., organism present at a concentration  10⁴ cfu/g tissue in an animal with histologic evidence of pneumonia). When using the usual discriminative values for PSB and BAL ( 10³ cfu/ml and  10⁴ cfu/ml, respectively), only 37% of the PSB specimens and 50% of the BAL specimens correctly established the microbiologic diagnosis of pneumonia. In contrast, with a discriminative value of 10⁵ cfu/ml, 94% of QEA specimens correctly established the microbiologic diagnosis of pneumonia. The bronchoscopic specimens that missed the microbiologic diagnosis either did not grow the causative organisms or recovered them, but at a concentration below the discriminative thresholds. Again, the uneven distribution of bacterial concentration for each single species may account for these findings. Although bacterial concentrations of single bacterial species recovered from matched left and right segments or from immediately adjacent segments significantly correlated, there was a rather large number of cases where one segment yielded high bacterial counts whereas the matched contralateral segment or even the immediately adjacent segment yielded low counts of the bacterial species or even was sterile. Similar observations were made by Chastre (4) when comparing bacterial concentrations of single bacterial species recovered from the anterior segment of the left upper lobe and the anterior segment of the left lower lobe. Therefore it is easy to understand why sampling a limited area of the lung, as done by PSB or BAL, may miss a causative organism.

Previous studies in humans already pointed out some of the technical difficulties encountered when using bronchoscopic techniques such as PSB and BAL for diagnosing VAP (21, 22). Recently, Sanchez-Nieto and coworkers questioned the real impact of these diagnostic techniques on patient management and outcome (23). The present study in a clinically relevant model of VAP provides additional insights into the peculiar histologic and bacteriologic features of VAP and thereby helps to understand why techniques such as PSB and BAL, combining quantitative cultures with the use of discriminative thresholds, cannot reliably recognize pneumonia and identify the causative organisms. For clinical practice we believe that no technique confidently helps in recognizing pneumonia in mechanically ventilated patients. With respect to bacterial diagnosis, QEA seems to be the best technique to identify the causative organisms in patients suffering VAP.