OBJECTIVE - To assess the incidence of fatal pulmonary embolism (FPE), the accuracy of clinical diagnosis, and the profile of patients who suffered an FPE in a tertiary University Hospital. METHODS - Analysis of the records of 3,890 autopsies performed at the Department of General Pathology from January 1980 to December 1990. RESULTS - Among the 3,980 autopsies, 109 were cases of clinically suspected FPE; of these, 28 cases of FPE were confirmed. FPE accounted for 114 deaths, with clinical suspicion in 28 cases. The incidence of FPE was 2.86%. No difference in sex distribution was noted. Patients in the 6th decade of life were most affected. The following conditions were more commonly related to FPE: neoplasias (20%) and heart failure (18.5%). The conditions most commonly misdiagnosed as FPE were pulmonary edema (16%), pneumonia (15%) and myocardial infarction (10%). The clinical diagnosis of FPE showed a sensitivity of 25.6%, a specificity of 97.9%, and an accuracy of 95.6%. CONCLUSION - The diagnosis of pulmonary embolism made on clinical grounds still has considerable limitations.