Thirteen RCTs (n=2,392) were included.
Eleven RCTs described methods of randomisation while eight described adequate allocation concealment.
Mortality and treatment failure.
There was no significant difference between empirical anti- Gram-positive antibiotics and the control in all-cause mortality (RR 0.86, 95% CI: 0.58, 1.26; 7 RCTs with 852 patients) or overall treatment failure (RR 1.00, 95% CI: 0.79, 1.27; 6 RCTs with 943 patients). No statistically significant heterogeneity was found. The results were similar for data from studies with adequate allocation concealment and studies with intention-to-treat data. There were significantly more treatment failures associated with treatment modifications in the control arm than with empirical anti-Gram-positive antibiotic treatment (RR 0.70, 95% CI: 0.61, 0.80; 5 RCTs with 1,178 patients).
Superinfections.
Compared with the control, glycopeptides significantly reduced bacterial superinfections (RR 0.3, 95% CI: 0.24, 0.59; 8 RCTs with 1,628 patients) and Gram-positive superinfections (RR 0.21, 95% CI: 0.11, 0.37). There was no significant difference between treatments for fungal infections.
Adverse effects.
Adverse effects were significantly more common with empirical anti-Gram-positive antibiotic treatment than with the control (RR 1.79, 95% CI: 1.55, 2.08; 8 RCTs with 1,546 patients). The most common adverse effects were dermatological (RR 2.31, 95% CI: 1.47, 3.63; 7 RCTs with 1,526 patients). There was no significant difference in nephrotoxicity for all antibiotics compared with a control (RR 1.28, 95% CI: 0.96, 1.70), but nephrotoxicity was significantly more common with glycopeptides than with a control (RR 1.43, 95% CI: 1.06, 1.94).
The funnel plot showed no evidence of publication bias.
Other results were reported.