Six studies were included: three retrospective (2,013 paired blood cultures) and three prospective (664 paired blood cultures). The lag time between the paired tests ranged from simultaneous sampling to four hours. All prospective samples discarded the initial blood drawn. Blood volume ranged from 3 mL to 10 mL.
Culture obtained from intravascular catheters had statistically significant better sensitivity (OR 1.85, 95% CI: 1.14, 2.99, range of sensitivities was 82.4 per cent to 94.8 per cent for central vascular catheters and 64.4 per cent to 95.4 per cent for peripheral venipuncture) and improved, but not statistically significantly so, negative predictive values (OR 1.55, 95% CI: 0.999, 2.39, range of negative predictive values was 97.6 per cent to 99.0 per cent for central vascular catheters and 95.5 per cent to 99.1 per cent for peripheral venipuncture) for diagnosing bacteraemia than peripheral venipuncture. Results from a fixed-effect model were reported.
Specificity was significantly lower when cultures were obtained from intravascular catheters (OR 0.33, 95% CI: 0.18, 0.59, range of specificities was 80.6 per cent to 95.6 per cent for central vascular catheters and 95.9 per cent to 98.4 per cent for peripheral venipuncture) as was positive predictive value (OR 0.41, 95% CI: 0.23, 0.76, range of positive predictive values was 17.2 per cent to 63.9 per cent for central vascular catheters and 66.7 per cent to 85.4 per cent for peripheral venipuncture).
Results from a random-effects model were reported. Results for the tests for heterogeneity were not reported. A decision analysis tree was presented to illustrate the numbers that would be appropriately and inappropriately treated if 1,000 blood cultures were sampled. The analysis tree showed that for every one additional patient diagnosed with true bacteraemia using blood drawn from intravascular catheters, more than seven would receive antibiotics inappropriately.