Fiberoptic bronchoscopy (FOB) has been reported to have a high diagnostic yield and to be safe in BMT patients with pulmonary infiltrates. At our institution, BMT patients with respiratory symptoms and/or pulmonary infiltrates had a thoracic CT and bronchoalveolar lavage (BAL). Transbronchial biopsy (TBBx) was considered if the platelet count could be raised to >30 x 10(9)/l. From March 1993 to August 1995, 52 patients had 68 FOBs (42 BAL + TBBx, 26 BAL only) for 60 episodes of clinical pneumonia. Patients' characteristics were: 38 males, mean age 42 years, and 39 allogeneic BMTs. Of the 68 FOBs, 47 were performed to evaluate diffuse infiltrates, 10 were done on mechanically ventilated patients, and 50 of the FOBs were preceded by a platelet transfusion. Thirty-one percent of FOBs (21 FOBs, 19 patients) were diagnostic. Twenty-four percent of FOBs (11 diagnostic FOBs, six nondiagnostic FOBs) changed therapy. Ten complications (15%) occurred in 10 FOBs (five acute respiratory failure, three pneumothoraces, one nose bleed, one death). Hospital and 6-month survival based on episodes of clinical pneumonia were 47 and 32%, respectively. Patients who had a diagnostic FOB or a FOB that changed treatment did not have better hospital or 6-month survival compared to patients who had FOBs that were nondiagnostic or did not change treatment. FOB in our BMT patient population, had a low diagnostic yield (31%), infrequently changed treatment (24%), a significant complication rate (15%) and was not associated with improved patient survival. The role of routine diagnostic FOB in BMT patients with pulmonary infiltrates and/or respiratory symptoms should be reevaluated.