In intensive care units, a large proportion of antibiotics are prescribed for presumed episodes of ventilator-associated pneumonia (VAP). VAP is usually diagnosed on a combination of clinical, radiographic, and microbiologic criteria with a high sensitivity but low specificity for VAP. As a result, patients may receive antibiotics unnecessarily. Specificity can be increased by the addition of quantitative cultures of samples of protected specimen brush (PSB) and bronchoalveolar lavage (BAL) to the diagnostic criteria. We prospectively analyzed the effects of implementation of PSB and BAL in the diagnosis of VAP on antibiotic prescription. PSB and/or BAL were performed in patients who fulfilled the clinical, radiographic, and microbiologic criteria for VAP. Based on quantitative cultures of PSB and/or BAL, patients were categorized into three groups: VAP microbiologically proven (Group 1; n = 72); clinical suspicion of VAP not confirmed microbiologically (Group 2; n = 66); and patients in whom bronchoscopy could not be performed (Group 3; n = 17). In Group 1, antibiotic therapy was instituted empirically in 40 patients (56%) (Group 1a) and after obtaining culture results in the other 32 patients (Group lb). Adjustment of therapy, based on culture results, occurred in 14 (35%) patients in Group la. In Group 2 empiric therapy was instituted in 34 (52%) patients (Group 2a) and dIscontinued within 48 h in 17 of them (50%). In Group 3, 17 (100%) patients were treated with antibiotics. Among the 66 patients in whom a clinical suspicion of VAP was not confirmed, only 18 (27%) were treated with antibiotics, and antibiotic therapy was withheld in 48 (35%) of 138 patients who underwent bronchoscopy. Withholding of antibiotic therapy had no negative effect on the recurrence of a clinical suspicion of VAP or on mortality rates. We conclude that addition of bronchoscopic techniques to the criteria for VAP may help to reduce antibiotic use. However, the definite benefits and cost-effectiveness of these techniques should be analyzed in a randomized study.