AIM: To compare the results of DNA amplification by the polymerase chain reaction (PCR) with immunofluorescence staining for detecting Pneumocystis carinii in bronchoalveolar lavage specimens taken from symptomatic HIV seropositive patients with suspected P carinii pneumonia (PCP). METHODS: Bronchoalveolar lavage specimens were obtained from 28 symptomatic HIV seropositive patients. Specimens were examined for P carinii using immunofluorescence, and by DNA amplification with PCR to obtain results on gel electrophoresis (gel) and a more sensitive Southern hybridisation (blot) technique. Specimens positive by immunofluorescence and gel electrophoresis were serially diluted to a 10(-6) concentration and each dilution strength tested for P carinii using PCR to compare quantitatively immunofluorescence with PCR. RESULTS: Of the 28 specimens analysed, 18 were negative for P carinii by both immunofluorescence and PCR, two were positive only by the blot technique of PCR, four were equivocally positive and four unequivocally positive by immunofluorescence. Three of the four equivocally positive patients tested by immunofluorescence were negative for P carinii by PCR, although one was positive by PCR (blot) technique. This patient had clinically confirmed PCP. Of the four unequivocally positive patients tested by immunofluorescence, three were gel and blot positive by PCR and had PCP clinically, but one was negative by both gel and blot techniques, although the patient certainly had PCP on clinical grounds. This patient had received nine days of treatment with high dose co-trimoxazole before bronchoalveolar lavage specimens were obtained. The three specimens positive by gel and blot techniques remained gel positive down to dilutions of between 10(-4) and 10(-6). CONCLUSIONS: PCR results may become negative soon after starting treatment for PCP. Specimens should therefore be taken before, or soon after, starting treatment. PCR seems to be between 10(4) and 10(6) times more sensitive than immunofluorescence.