A consecutive series of 1,390 primary total knee arthroplasty (TKA) procedures (1,201 patients, 1,600 arthroplasties) performed between January 1980 and July 1994 were reviewed to establish the incidence of death from pulmonary embolism (PE). Nine hundred twenty-three bi- or tricompartment TKAs and 467 unicompartment TKAs were performed as one-stage procedures. Chemical thromboprophylaxis was used only in high-risk cases in which there was a history of previous thromboembolism or obesity. There were no deaths from PE after unicompartment arthroplasty procedures. Autopsy confirmed PE as the cause of death in 2 patients following bi- and tricompartment TKAs (0.22%; 95% confidence interval [CI], 0.03-0.8%). The incidence was higher for one-stage bilateral TKA as 1 of the autopsy-confirmed deaths occurred in this group 0.7% (95% CI, 0.02-3.78%). Two other deaths were certified without postmortem examination (pneumonia and myocardial infarction in each case). As PE could not be ruled out as the cause of death in the latter 2 cases, these were considered as possible PE deaths to provide the maximum possible death rate that could result. Thus, the maximum possible incidence of fatal PE after TKA without routine use of chemical anticoagulation was 0.4% (95% CI 0.1-1.1%). It is concluded that the risk of fatal PE after unilateral TKA and unicompartment knee arthroplasty is low. The risk of clinical, nonfatal thromboembolic events, which might themselves warrant prophylaxis, was not quantified in this article.