The different methods for the diagnosis of bacterial pneumonia in mechanically ventilated patients are all imperfect because of their lack of sensitivity or specificity. At the present time, the reference method is the pathologic examination and/or culture of lung biopsy specimens, but it cannot be performed routinely in intubated patients. The PSB is considered as one of the reference methods used for the diagnosis of ventilator-associated pneumonia (VAP). Recently, BAL was proposed in nonventilated-patients and then validated in ventilated baboons.
However, the delay of diagnosis due to time of the bacterial growth is 24 to 72 h, compromising rapid institution of appropriate antibiotic therapy in critically ill patients.
Recent studies have shown the usefulness of microscopic examination of the BAL fluid for early diagnosis of VAP. Gram’s and Giemsa stains identify most of the microorganisms that grow at significant concentrations but their validity is a subject of controversy. In the study by Rello et al, the diagnosis of VAP was confirmed only in 46% of the cases. Furthermore, these tests do not distinguish colonization and infection. In our study, sensitivity and specificity of presence of bacteria on Gram’s stain for the diagnosis of bacterial pneumonia were, respectively, 92% and 76.5%. However, in terms of qualitative agreement, the concordance was poor: in 44% of the cases, the Gram’s stain was partially right (for example, gram-negative bacilli on Gram’s stain and S aureus plus P aeruginosa on PSB cultures) or totally false (gram-negative bacilli on Gram’s stain and gram-positive stain on PSB cultures). Gram’s stains seem to be useful to start an empiric antimicrobial therapy but not for the choice of appropriate antibiotics. The diagnostic accuracy of a Gram’s stain is at least as good as the examination of intracellular bacteria in our series of patients. Unfortunately, few data are available on the usefulness of Gram’s stain in BAL in case of VAP.
Chastre and coworkers related the usefulness of BAL cytologic study with search of intracellular microorganisms to assert the early diagnosis of nosocomial pneumonia. With a threshold of 7% of infected cells, in a study in 61 patients (14 had pneumonia), there was a good prediction of infection (sensitivity, 86%; and specificity, 96%). The cutoff values differ between investigators.
Papazian et al reported the presence or absence of intracellular organisms (ICs). In this study, the presence of neutrophils containing bacteria was highly specific (98.3%) but the lack of sensitivity (19.2%) made this parameter unreliable for the diagnosis of VAP. Pugin et al detected PMNs with ICs in 11 of the 15 patients with infection (73%) and no ICs in the 25 patients without bacterial pneumonia. Meduri and colleagues, using a protected transbronchoscopic balloon-tipped catheter, found ICs in more than 2% of the alveolar cells in 11 of the 13 patients with bacterial pneumonia (84%) and in none of the 33 patients without bacterial pneumonia. Sole-Violan and colleagues used a threshold of 4% with a sensitivity of 62.5% and a specificity of 100% in a study with 33 BALs. In a recent study on 80 BALs (28 pneumonias), Aubas and coworkers, using an ROC curve, attempted to define a threshold for ICs in pneumonia, but the AUC was 0.718 and no cutoff value could be defined. By using a cutoff value of 5%, the sensitivity was 39.3% and the specificity was 98%.
We performed a study on 163 BAL samples (in 132 patients): 56 of the cases had confirmed pneumonia and 107 cases constituted the control group. The count of ICs was established on 100 cells (the diagnostic assessment was not performed in a blinded fashion). An ROC curve showed an AUC of 0.888. Sensitivity was estimated at 84% and specificity at 80%, positive predictive value was 69%, and negative predictive value was 90%.
Dotson and Pingleton suggested that prior and current antibiotic therapy reduced recovery of ICs from BAL and reduced accuracy of IC count for the diagnosis of VAP. Conversely, Timsit et al concluded that previous antibiotic therapy received to treat an earlier septic episode unrelated to suspected pneumonia did not affect the diagnostic yield of PSB and BAL (quantitative cultures and direct examination of centrifuged BAL fluid). Our data also suggest that even in case of prior antimicrobial chemotherapy (except recent change <72 h), the quantification of ICs from BAL remains useful. Among the group receiving antibiotics, the sensitivity and the specificity were, respectively, 82% and 77%; among the group without antibiotics, 85% and 83%.
In our study, the incidence of S aureus pneumonia was high and confirmed that S aureus has emerged as the second most common cause of nosocomial pneumonias. The National Nosocomial Infection Study (NNIS) provides the largest database describing pathogens isolated from both ventilated and nonventilated patients with nosocomial pneumonia. Based on NNIS data from 1986 to 1989, the most frequently implicated pathogen is P aeruginosa (17%), followed by S aureus (16%), Enterobacter species (11%), Klebsiella pneumoniae (7%), and Escherichia coli (6%)
The criteria for entry into the study included all four of the following: fever, purulent sputum, leukocytosis, and radiographic infiltrates. The sensitivity of the technique, and therefore the optimum ROC curve cutoff, may be different in patients who lack one or more of these entry criteria.
The count of total cells was significantly higher in control patients. The SDs are huge; however, the percentage of recovered instilled fluid may be very different according to the patient status. In our opinion, the count of total cells is particularly useful to assess the quality of BAL.
The optimal technique for the diagnosis of nosocomial pneumonia remains unclear. The most recent studies using histologic features as the gold standard are divergent.
For Chastre et al, PSB quantitative culture results strongly correlated with lung tissue values. Using a cutoff of 103 cfu/mL, they found a sensitivity of 82% and a specificity of 89%. In this last study, the authors report that the percentage of BAL cells containing intracellular bacteria was closely correlated with the total number of bacteria obtained from corresponding lung sample. Marquette et al found, in 28 mechanically ventilated patients, a sensitivity of 57% and a specificity of 88% for quantitative cultures of PSB. The presence of intracellular organisms in BAL had a 36.8% sensitivity and 100% specificity in establishing the diagnosis of pneumonia. They found that the use of a diagnostic threshold for the percentage of intracellular organisms was unnecessary, since the presence of intracellular organisms, whatever the percentage of ICs, was always associated with pneumonia. However, Chastre et al used histologic features and lung biopsy quantitative cultures, whereas Marquette et al used only histologic examination as the gold standard. The gold standard to evaluate diagnosis techniques remains a subject of controversy.