The objectives of the study were to determine the agreement between the protected specimen brush technique (PSB) with quantitative cultures and endotracheal aspirates (EA) with quantitative cultures when using increasing interpretative cutoff points and to investigate the respective operating characteristics for the diagnosis of pneumonia of PSB and EA when using quantitative cultures. Consecutive sampling of respiratory secretions using these two techniques was conducted in the respiratory intensive care units in 52 mechanically ventilated patients with clinical and radiologic suspicion of pneumonia. Quantitative bacterial cultures of PSB and EA samples were obtained. The 10(6) cfu/ml cutoff point was the most accurate diagnostic threshold for the EA technique. When using this threshold, there was a high level of agreement (84.6%) between PSB and EA results. Among the few discrepancies, the EA result was always indicative of pneumonia, whereas the PSB result was nonindicative, thus permitting us to classify correctly five patients in whom pneumonia would have been erroneously excluded on the basis of the sole result of PSB. Conversely, there was no case where the PSB result was indicative of pneumonia when the EA result (at the 10(6) cfu/ml level) was not. The operating characteristics of the PSB technique for the diagnosis of pneumonia were in accordance with previously published studies. The operating characteristics of the EA technique (when taking the 10(6) cfu/ml of respiratory secretions as the interpretative cutoff point) compared favorably with those of the PSB technique. Diagnostic accuracy rates were similar. The specificity of EA was somewhat lower (83 versus 96%), but the sensitivity was higher (82 versus 64%).(ABSTRACT TRUNCATED AT 250 WORDS)