The predictive value of procalcitonin serum levels to detect or rule out bacteremia was investigated prospectively in a case-control study with 200 hospitalized adults from whom blood samples were taken for culture. Fifty bacteremic patients (cases) had higher procalcitonin serum levels than the 150 controls with sterile blood cultures (11.7 vs. 0.7 ng/ml; P=0.0001), a difference that remained significant after controlling for potential confounders in multivariate analysis. At cut-off values of 0.5 and 0.2 ng/ml, the sensitivity of procalcitonin was 56 and 92%, and the specificity was 83 and 43%, respectively. These results yielded low positive (22 and 12%) and high negative predictive values (96 and 99%), reflecting primarily the low prevalence of bacteremia among patients who undergo blood cultures in hospitals (low pretest probability). Although caution is mandatory when using such markers at the individual level, procalcitonin, possibly together with other parameters, could nonetheless prove useful in future studies to rapidly rule out bacteremia.