The aim of this prospective study was to compare the results obtained with the non-protected lower respiratory tract secretions samples (LRS) with the protected specimen brushes (PSB) performed through a fiberoptic bronchoscope in mechanically ventilated patients, when pneumonia was suspected. The diagnosis of pneumonia was ultimately made at the end of the hospitalisation, in a double-blind manner by 2 members of the medical staff not aware of the bacteriologic results of LRS and PSB. LRS and PSB were performed in 24 patients. PSB culture was considered as positive at a level of 10(3) colony-forming units per milliliter (cfu/ml) microorganisms. Twenty-five samples from 24 patients were divided as follows: (1) LRS (-) and PSB (-) 5 samples: the clinical diagnosis of pneumonia was never established. (2) LRS (+) and PSB (+) 10 samples: the clinical diagnosis of pneumonia was always established, 2 microorganisms were involved 4 times and 1 microorganism 6 times. (3) LRS (+) and PSB (-) 10 samples: the clinical diagnosis pneumonia was retained in 3 with the possibility of false negative PSB. We conclude that (1) a negative LRS eliminated the diagnosis of pneumonia without PSB; (2) a positive LRS was not sufficient to diagnose pneumonia since PSB was negative in 50% of all LRS (+) cases; (3) the possibility of a false negative PSB must be kept in mind particularly in patients previously treated with antibiotics; (4) 2 microorganisms may be responsible for the pneumonia if the previously determined, as significant, bacteriological count (greater than or equal to 10(3) cfu/ml) appears to be accurate.