To assess the accuracy of clinical parameters for the diagnosis of ventilator-associated (VA) pneumonia, as well as the diagnostic value of several invasive techniques, such as protected specimen brush (PSB), bronchoalveolar lavage (BAL), fiberoptic bronchial aspirates (FBAS), and percutaneous lung needle aspiration (PLNA), we compared the results of these techniques with the histopathology of immediate postmortem pulmonary biopsies, considered the "gold standard" reference test. We studied 30 mechanically ventilated patients (age 52 +/- 21 yr; mechanical ventilation period 9 +/- 7 days) who died in an intensive care unit. All patients received prior antibiotic treatment. The following procedures were performed immediately after death: bilateral PSB, BAL, FBAS, and PLNA, as well as bilateral minithoracotomies to obtain pulmonary biopsies as close as possible to the area sampled with the other techniques. According to the histopathology 18 patients had pneumonia and 12 did not. The presence of fever (sensitivity 55%, specificity 58%), purulent secretions (sensitivity 83%, specificity 33%), and chest radiograph infiltrates (sensitivity 78%, specificity 42%) could not differentiate in all instances presence from absence of pneumonia. Quantitative bacterial cultures of lung biopsies using 10(3) cfu/g as a cutoff point had low sensitivity (40%) and low specificity (45%) and could not differentiate the histologic absence or presence of pneumonia. Considering the histopathology of pulmonary biopsies as a gold standard, we found the following sensitivities for PSB, BAL, FBAS, and PLNA: 36, 50, 44, and 25%. The specificities were 50, 45, 48, and 79%, respectively. The sensitivities and specificities of different invasive techniques are much lower than those reported in clinical studies.(ABSTRACT TRUNCATED AT 250 WORDS)