The specificity of the different techniques used to diagnose ventilator-associated pneumonia is still a matter of controversy. To investigate the specificity of endotracheal aspiration (EA), protected specimen brush (PSB), and bronchoalveolar lavage (BAL) quantitative cultures, we studied 27 consecutive mechanically ventilated (MV) patients (> 72 h) without clinical or radiographic evidence of pulmonary infection. Comparing different thresholds for quantitative cultures (from 10(3) through 10(6) CFU/ml), the lowest rate of false positive results was obtained using 10(6) for EA, 10(5) for PSB, and 10(6) for BAL. Using 10(6) CFU/ml for EA, 10(4) CFU/ml for PSB, and 10(5) CFU/ml for BAL as cutoff points, we obtained the following specificities: 85, 85, and 78% for the three techniques, respectively. A bacterial index of 8 was the best threshold to get a low percentage of false positive results for all techniques except for EA (0% for PSB and 12% for BAL). There were reasonable qualitative agreements (PSB versus EA = 58%; BAL versus EA = 69%; and PSB versus BAL = 62%) and poor quantitative correlations between concomitantly isolated microorganisms from the three types of samples. Quantitative cultures of EA, PSB, and BAL may show a considerable percentage of false positive results at the respective cutoff points usually accepted. Increasing the thresholds for quantitative cultures, albeit loosing sensitivity, may rule out better the absence of pulmonary infection in MV patients.