The value of a new technique of protected bronchoalveolar lavage not requiring bronchoscopy was prospectively evaluated for the diagnosis of nosocomial pneumonia in two groups of critically ill patients. The control group was comprised of 29 patients free of any pulmonary disease whose lungs were ventilated for a mean time of 14 +/- 9 days (mean +/- SD). The pneumonia group was comprised of 30 patients with histologically proven nosocomial pneumonia whose lungs were ventilated for a mean time of 11 +/- 8 days. All patients of the pneumonia group died, and postmortem lung biopsies were taken for bacteriologic and pathologic examination. Twice a week in the control group or within the day preceding death in the pneumonia group, distal bronchial samples were obtained by a technique of protected bronchoalveolar lavage performed at the bedside. Fifty-one distal bronchial samples were bacteriologically analyzed in the control group and 30 in the pneumonia group. The sensitivity of a positive protected bronchoalveolar lavage for diagnosing nosocomial pneumonia was 80%, whereas the specificity was 66%. In 73% of the patients of the pneumonia group, the microorganisms isolated in the protected bronchoalveolar lavage and in the lung culture were partially (16%) or completely in agreement (57%). Among the 43 microorganisms isolated in the lung cultures, 74% were recovered by the protected bronchoalveolar lavage technique. This study shows that the protected bronchoalveolar lavage is an accurate technique for the diagnosis of nosocomial pneumonia. Because the technique is simple, noninvasive, easily repeatable at the bedside, and enables gram staining, it represents an attractive alternative to the fiberoptic bronchoscopy technique using a plugged double-sheathed brush.