To determine the usefulness of samples obtained by bronchoalveolar lavage (BAL) in establishing the diagnosis of ventilator-associated pneumonia, quantitative cultures of BAL and protected specimen brush (PSB) samples obtained via fiberoptic bronchoscope were compared in 42 patients with suspected ventilator-associated pneumonia. Direct examination of BAL fluid was also used to identify cells with intracellular organisms. Ventilator-associated pneumonia was diagnosed in 18 patients; a total of 39 microorganisms were recovered from BAL fluid and 29 from PSB specimens. Cultures of 21 BAL and 23 PSB specimens were sterile. Quantitative BAL and PSB cultures coincided in 76% of cases. Sterile BAL and PSB cultures agreed in 87% of cases. Cultures were completely discordant in only three cases. The sensitivity of BAL for diagnosis of ventilator-associated pneumonia using bacterial counts of > or = 10(4) cfu/ml was 89%, and specificity was 100%. In 14 of the 18 patients with ventilator-associated pneumonia, the percentage of cells containing intracellular organisms in specimens recovered by BAL was 11.6% versus 0.45% in patients without pneumonia (p < 0.05). In the remaining four patients, all of whom had Pseudomonas aeruginosa pneumonia, no intracellular organisms could be detected. Using a cut-off point of > or = 5% of cells with intracellular organisms, the sensitivity and specificity for the early diagnosis of ventilator-associated pneumonia was 67% and 96%, respectively. The results confirm the usefulness of the quantitative BAL culture (with a cut-off at 10(4) cfu/ml) for the diagnosis of ventilator-associated pneumonia. The identification of intracellular organisms in BAL fluid is a good early indicator of pneumonia, but the sensitivity of this technique may be lower for Pseudomonas aeruginosa infections.