One hundred forty-four cancer patients harboring a central venous catheter (CVC) were prospectively investigated to assess the relationship between hub culture, clinical assessment of sepsis before removal, and CVC sepsis. In 22 patients, the CVC was removed because of clinical assessment of catheter sepsis expressed by the staff prior to the removal. For each CVC removal, peripheral blood (qualitative method), hub, and CVC tip (quantitative method) cultures were performed. Clinical sepsis (disappearance of fever after CVC removal) was observed in 13 patients, microbiologic "sepsis" (identification of the same microorganisms on the CVC tip and in the peripheral blood) in seven patients, and clinical and/or microbiologic sepsis in 16 patients. Staphylococcus epidermidis was the microorganism most frequently identified. Hub culture was negative in 48% and positive for a low number and a high number of colonies in 35% and 17%, respectively. The predictive value of hub culture was 96% when testing negative and 8% and 37% (p = 3 x 10(-3)) when testing positive for a low and a high number of colonies, respectively. Predictive values of clinical assessment were 55% if positive and 97% if negative. Combining hub cultures and clinical assessment, the risk of sepsis varied from 2% with both evaluations negative to 89% in the case of positive clinical assessment associated with positive high-count hub. Inasmuch as the CVCs used have a disposable hub, it is possible to have an accurate diagnosis of CVC sepsis without removing the CVC.