OBJECTIVE The concept of the protected bronchoalveolar lavage (PBAL) is to improve the diagnostic yield in bacterial pneumonia by unifying the high sensitivity of bronchoalveolar lavage (BAL) and specificity of the protected specimen brush (PSB). HIV-infected patients have been shown to have a high incidence of bacterial pneumonia as well as bacterial colonisation of the tracheobronchial tree. We therefore studied the value of PBAL in this population. METHODS During a period of twelve months 40 episodes in 36 patients with symptoms suggestive of pneumonia were investigated retrospectively. In all cases without infiltrates on chest radiograph a CT-scan of the chest was performed. Patients without infiltrates also in CT-scan served as controls. Bronchoscopic investigation included a PSB and a PBAL in the same lung segment most prominently affected. Microbiological workup was performed for bacterial agents, mycobacteria, fungi, viruses and parasites. Quantitative cultures for bacteria were considered significant in case of > or equal to 10(3) cfu/ml in PSB and > or equal to 10(4) cfu/ml in PBAL. RESULTS 32 episodes in 28 patients were identified as pneumonia. A definite diagnosis could be established in 19/32 (59%) of cases. Bacteria accounted for 10/19 (53%). Pneumocystis carinii for 9/19 (47%) of cases including one case that revealed mixed infection with Streptococcus pneumoniae and Pneumocystis carinii. Another pneumonia was due to Aspergillus fumigatus. The sensitivity for bacterial pneumonia was 44% for PSB and 56% for PBAL, the specificity 100%. The overall diagnostic accuracy was 60% and 68%, respectively. The yield for Pneumocystis carinii was 8/9 (89%). CONCLUSIONS PBAL as compared to PSB had a superior diagnostic yield for bacterial pneumonia. As PBAL additionally conserved the yield of BAL reported for Pneumocystis carinii, it may represent a rational diagnostic technique for pneumonia in HIV-infected patients.