The aim of this study was to evaluate the diagnostic efficacy of protected bronchoalveolar lavage (PBAL) in ventilator-associated pneumonia (VAP), and to determine the effect of antibiotic therapy on its microbiological and cytological results. We prospectively studied 102 episodes of suspected VAP in 93 patients. Subsequent follow-up confirmed VAP in 35 of the 102 (34%) cases. In 55 of the 102 (55%) VAP was ruled out, and the diagnosis remained undetermined in 12 of the 102 (12%) episodes. In the VAP group, 30 of the 35 (86%) PBAL (> or = 10(4) colony-forming units (cfu)-mL-1) cultures were positive. In the non-VAP group, 5 of the 55 (9%) PBAL cultures were positive. A Giemsa stain of PBAL samples was performed in 32 of the 35 cases of VAP. Intracellular organisms (ICO) were found in 24 of the 32 (75%) cases. Seven of the other eight cases without evidence of ICO were already on antibiotics. In the non-VAP group, ICO were present in only 1 out of 55 (2%) cases. The mean ICO was significantly higher in the group who had not received antibiotics when compared with those patients previously treated for less than 48 h (p < or = 0.01) and those treated for more than 48 h (p = 0.009). The sensitivity of protected bronchoalveolar lavage quantitative cultures was 87% and the specificity 91%. The sensitivity of cytological analysis for intracellular organisms was 75% and the specificity 98%. According to our results, if the patient is already on antibiotics, the direct examination of protected bronchoalveolar lavage fluid is less reliable, although still helpful.