In the United States, therapeutic plasma exchange (TPE) is both performed and requested by a wide range of services, often on an empiric basis (before a diagnosis is established). Whether empiric therapy is beneficial has not been established. Patients were identified from an electronic procedure log that included those patients who received plasmapheresis at Walter Reed Army Medical Center from 1996 to 2003. The clinical indications, referring service, and outcomes (including deaths) that occurred were tabulated. Between March 1997 and August 2003, 568 TPE treatments were performed in 54 patients. The majority of the diagnoses were either neurologic (48%) or hematologic (37%). Thirty-three patients (61%) received TPE for a Category I indication. Twelve cases were performed empirically (without an established diagnosis) at the request of the referring service, most (7) performed for presumed thrombotic thrombocytopenic purpura (TTP). Almost 80% of patients required central venous catheters for treatment. Twelve patients (22%) experienced a major complication including death, and six patients (11%) died. Of the patients who died, 5 (83%) were treated empirically versus one death (17%) among patients not treated empirically, P < 0.001 by Chi Square. Only one of the seven patients treated empirically for TTP died, however. In logistic regression analysis, empiric treatment was the only factor independently associated with death, adjusted odds ratio, 34.2, 95% CI, 3.4, 334.8, P ΒΌ 0.003. The most common indication for TPE was neurological disease, which also accounted for the highest proportion of complications. With the exception of presumed TTP, performing TPE in the absence of a confirmed diagnosis was not beneficial.