To determine the primary sources and secondary complications of Staphylococcus aureus bacteremia (SAB) in cancer patients, as well as predictors of outcome in cancer patients with SAB. Fifty-two patients at Duke University Medical Center met entry criteria between September 1994 and December 1996 for this prospective cohort study involving hospitalized nonneutropenic adult cancer patients with SAB. All subjects were observed throughout initial hospitalization and were evaluated again at 6 and 12 weeks or until death. SAB was intravascular device-related in 42%, tissue infection-related (TIR) in 44%, and unidentifiable focus-related (UFR) in 13%. Seventeen patients (33%) were found to have metastatic infections or conditions, with eight (15%) developing infectious endocarditis (IE). Patients with TIR bacteremia were less likely than other patients to develop IE (4% v 24%, P =.06). The overall mortality rate was 38%, the SAB-related mortality rate was 15%, and the rate of SAB relapse was 12%. Methicillin resistance was not associated with adverse outcome. Inability to identify a point of entry (UFR bacteremia), however, was associated with a higher overall mortality rate (100% v 24%, P =.0006). Furthermore, a 72-hour surveillance blood culture positive for organisms was associated with an increased incidence of IE (P =.0006), metastatic infections or conditions (P =.0002), SAB relapse (P =.038), and SAB-related death (P =.038). SAB in cancer patients is associated with significant morbidity from frequent metastatic infections or conditions including IE, as well as considerable mortality. Unknown initial infection site and 72-hour surveillance cultures positive for organisms were predictive of a complicated course and poor final outcome.