The purpose of this prospective postmortem study was to assess the diagnostic accuracy of bronchoscopic techniques (bronchoalveolar lavage [BAL] and protected specimen brush [PSB]) and nonbronchoscopic techniques (blind bronchial sampling [BBS] and mini-BAL) in the diagnosis of ventilator-associated pneumonia (VAP). The results of each technique were compared with histology and culture of lung tissue specimens obtained by surgical pneumonectomies in 38 patients who died after at least 72 h of mechanical ventilation. Histology was positive for VAP in 18 patients and negative in 20 patients. There were 12 definite VAP (positive histology and positive lung cultures) and 6 histologic VAP (positive histology and negative cultures). Clinical pulmonary infection score (CPIS) at a threshold of 6 achieved a sensitivity of 72% and a specificity of 85%. When the CPIS was combined with the logarithmic concentration of the predominant microorganism obtained from the BBS sample culture, specificity was increased to 95%, for a threshold of 10. Using 10(3) cfu/ml as the threshold of positivity for cultures obtained with PSB and mini-BAL samples and 10(4) cfu/ml for cultures obtained with BBS and BAL, the respective sensitivities of these techniques for definite VAP were 42, 67, 83, and 58%. The sensitivity of BBS was significantly higher than that of PSB (p < 0.05). The area under the receiver operator characteristic curve was significantly greater for BBS than PSB (p < 0.05). Given that it is more sensitive and noninvasive, BBS is preferable to PSB for the diagnosis of VAP.