The diagnostic use of procalcitonin for bacterial infections remains a matter of debate. Most studies have used ambiguous outcome measures such as sepsis instead of infection. We performed a systematic review and meta-analysis to investigate the diagnostic accuracy of procalcitonin for bacteraemia, a proven bloodstream infection. We searched all major databases from inception to June 2014 for original, English language, research articles that studied the diagnostic accuracy between procalcitonin and positive blood cultures in adult patients. We calculated the area under the summary receiver-operating characteristic (SROC) curves and pooled sensitivities and specificities. To minimize potential heterogeneity we performed subgroup analyses. In total, 58 of 1567 eligible studies were included in the meta-analysis and provided a total of 16,514 patients, of whom 3420 suffered from bacteraemia. In the overall analysis the area under the SROC curve was 0.79. The optimal and most widely used procalcitonin cut-off value was 0.5 ng/mL with a corresponding sensitivity of 76% and specificity of 69%. In subgroup analyses the lowest area under the SROC curve was found in immunocompromised/neutropenic patients (0.71), the highest area under the SROC curve was found in intensive-care patients (0.88), sensitivities ranging from 66 to 89% and specificities from 55 78%. In spite of study heterogeneity, procalcitonin had a fair diagnostic accuracy for bacteraemia in adult patients suspected of infection or sepsis. In particular low procalcitonin levels can be used to rule out the presence of bacteraemia. Further research is needed on the safety and efficacy of procalcitonin as a single diagnostic tool to avoid taking blood cultures.