A total of 34 randomised controlled trials (RCTs) were included (n=10,450 central venous catheters, range 50-777). The included trials were generally of poor quality, with allocation concealment described in eight trials, five trials were double blinded and three trials provided an intention-to-treat analysis.
Statistically significantly better outcomes were reported when comparing antimicrobial central venous catheters with standard central venous catheters for central venous catheter colonisation (OR 0.54, 95% CI: 0.43, 0.67, 29 trials) and for catheter-related bloodstream infections (OR 0.58, 95% CI: 0.45, 0.75, NNT 77, 28 trials).
Reduced colonisations with antimicrobial central venous catheters compared with standard central venous catheters for first-generation chlorhexidine-silver sulfadiazine were reported (OR 0.51, 95% CI: 0.42, 0.61, 13 trials). First generation chlorhexidine-silver sulfadiazine central venous catheters also reduced catheter-related bloodstream infections (OR 0.68, 95% CI: 0.47, 0.98, NNT 72). Second generation chlorhexidine-silver sulfadiazine central venous catheters (three trials) reduced central venous catheter colonisation (OR 0.39, 95% CI: 0.25, 0.60) but not catheter-related bloodstream infections.
Four trials assessed colonisation associated with minocycline-rifampicin central venous catheters with a reduction in colonisations compared with standard central venous catheters (OR 0.39, 95% CI: 0.27, 0.55). Five trials assessed the catheter-related bloodstream infection outcome (OR 0.29, 95% CI: 0.16, 0.52, NNT 21), showing reduced catheter-related bloodstream infection incidence with antimicrobial central venous catheters compared with standard central venous catheters.
Under sensitivity analysis, minor increases in effectiveness for catheter-related bloodstream infections were reported where trials were excluded that: allowed the use of more than one study central venous catheters per patient; allowed guide wire exchange; did not report the requirements of clinical symptoms; and drop-outs were greater than 15% or were not reported. Significant publication bias was not detected.