Twenty-four cancer patients with diffuse interstitial pneumonitis (DIP) were randomized to undergo an open lung biopsy (OLB) within 8 hours of presentation (12 patients) or to receive empiric antimicrobial therapy (ET) with trimethoprim-sulfamethoxazole (TMP-SMX) erythromycin for a minimum of 4 days (12 patients). Patients whose condition deteriorated underwent an OLB on day 4. Eight of 12 patients (67%) having OLB survived versus 10 of 12 (83%) receiving ET (P = .64). Morbidity occurred in nine of 12 (75%) having OLB versus eight of 12 (67%) receiving ET (P = 1.0). Concurrently, there were 14 additional cancer patients with DIP who were not randomized (nine refused, three had a coagulopathy contraindicating surgery, two were excluded by primary care physicians) and who were comparable demographically to the randomized group. Two received OLB and 12 ET. Combining the randomized and nonrandomized groups, eight of 14 (57%) having an initial OLB survived versus 18 of 24 (75%) of ET-treated patients (P2 = .19). Results of the OLB were seven Pneumocystis carinii pneumonia (PCP), five nonspecific pneumonitis (NSP), one cytomegalovirus, and one lymphoma. Results of OLB led to discontinuation of antibiotics in three patients. Of the 24 ET patients, eight failed to improve by day 4 and had an OLB. Results were two NSP, two PCP, two cancer, one blastomycosis, and one Candida pneumonia. Complications were seen in 10 of 14 (72%) initial OLB patients versus 14 of 24 (58%) patients on the ET arm (P = .65). When the complication rate between patients receiving only empiric antibiotics was compared with all patients having an OLB (initially or on day 4), the difference was greater in patients undergoing OLB (37% v 72%, respectively) (P2 = .14). ET with TMP-SMX plus erythromycin and broad spectrum antibiotics in granulocytopenic patients appeared to be as successful and potentially less toxic than an OLB in this study. Although the number of patients in this study was small, these data suggest that a trial of empiric antibiotic management may be reasonable in cancer patients presenting with DIP, especially if they are nonneutropenic.