Quantitative cultures of specimens obtained at fiberoptic bronchoscopy have been used to diagnose nosocomial pneumonia in research settings, but their clinical role remains controversial. We reviewed the literature comparing these culture techniques with other accepted methods to diagnose pneumonia in mechanically ventilated patients and extracted data to describe the receiver operator characteristics (ROC) of quantitative cultures of protected specimen brush (PSB) and bronchoalveolar lavage (BAL) samples. Analysis of ROCs reveals that these tests have a discriminating power comparable or superior to that of many widely accepted routinely used tests. Current data do not suggest that either culture technique offers an advantage over the other. Since benefits of antibiotic therapy of pneumonia and risks of treatment of noninfected critically ill patients are not well quantified, universally applicable recommendations for appropriate values to define an abnormal test result cannot logically be made. Multiple decision analytic tools show that values lower than those previously recommended are more appropriate in patients suspected of having pneumonia unless the risk of antibiotic therapy is judged to be extreme. On the basis of these findings, we suggest guidelines for clinicians' interpretation of PSB and BAL quantitative culture results.