The aetiological diagnosis of nosocomial pneumonia in intensive care unit (ICU) patients requires a valid, cheap and safe method. This method should be suitable for all mechanically-ventilated patients and all ICUs. The aim of this study was to assess the diagnostic yields of three methods: "blind" bronchial brushing (Accu-Cath)(protective specimen brush-nonbronchofibroscopic (PSB-non BF)); bronchofibroscopic protected specimen brushing (PSB-BF) and bronchoalveolar lavage (BAL).We prospectively studied the diagnostic values of the three methods as well as the agreement between microbiological results in 74 patients with 88 episodes of clinically suspected ventilator-associated pneumonia (VAP) and 22 control subjects. VAP episodes were also divided into those with (n=24) and without antibiotic pretreatment (n=64), and into those with (n=78) and without (n=10) right lower lobe infiltrates on chest radiography.No differences were found as regards the bacteriological yield of the three techniques. Furthermore, the rate of concordant results was high; 92% for PSB-BF and BAL; 84% for PSB-nonBF and BAL; 85% for PSB-nonBF and PSB-BF; and 85% for PSB-nonBF combined with both bronchoscopic techniques. The diagnostic yields in suspected VAP were 66, 59 and 56% for PSB-nonBF, PSB-BF and BAL, respectively.We conclude that "blind" bronchial brushing has similar accuracy to bronchoscopic techniques commonly used in the diagnosis of ventilator-associated pneumonia, constituting an interesting alternative in hospitals where fibreoptic bronchoscopy is not available.