We compared the efficacy of three techniques--minimal time to positivity (MTTP) of blood cultures (BCs), differential time to positivity (DTTP) of BCs obtained from the catheter and peripheral veins and the number of positive BCs--in predicting catheter involvement in patients with well-demonstrated catheter-related candidaemia (C-RC) and non-catheter-related candidaemia (NC-RC).C-RC was defined as isolation of the same Candida species from blood and catheter tip culture (≥15 cfu/plate). A ROC curve was created for each quantitative variable to determine the best cut-off for predicting C-RC.A total of 108 episodes of candidaemia were included (84 adults and 24 children; 67 C-RC and 41 NC-RC). These were caused mainly by C. albicans (49.1%) and C. parapsilosis (30.6%). The MTTP was significantly shorter in adult patients with C-RC than in those with NC-RC (29.8 vs. 36.8 hours; p 0.035), although no cut-off value provided acceptable accuracy. DTTP had high sensitivity but low specificity for predicting CRC. However, C-RC episodes had a significantly greater number of positive BCs than NC-RC episodes. The optimal cut-off for predicting C-RC was at least two positive BCs out of three, with the following validity values: sensitivity, 100%; specificity, 62.5%; positive predictive value, 83.3%; negative predictive value, 100%; accuracy, 87.0%.None of the tests evaluated allow a clear-cut prediction of C-RC and the criteria accepted for bacteraemia should not be automatically extrapolated to candidaemia. We found that a low number of positive BCs with Candida had a high negative predictive value for a catheter origin.
© 2012 The Authors Clinical Microbiology and Infection © 2012 European Society of Clinical Microbiology and Infectious Diseases.