Eight studies (n=1,877) were included: 5 RCTs (n=495) and 3 prospective comparative studies (n=1,382).
The mean quality score of the RCTs was 2.6, which was considered good. Seven of the studies were conducted pre-1995. The authors noted that current ICU antimicrobial resistance patterns differ considerably from those reported in the included studies, most of which isolated only a small proportion of methicillin-resistant staphylococci.
No statistically significant difference was found between the intervention group and controls receiving placebo or no therapy in either all-cause mortality (5 RCTs, n=495) or pneumonia-related mortality (2 RCTs, n=179). The inclusion of non-randomised studies (total n=1,837 for all-cause mortality and n=923 for pneumonia-related mortality) did not change these findings.
ICU-acquired pneumonia was significantly less common in the the intervention arm than in control groups (5 RCTs, n=495; OR 0.49, 95% CI: 0.32, 0.76). The inclusion of non-randomised studies barely altered the effect estimate (total n=1,837).
Rates of colonisation with Pseudomonas aeruginosa were similar in the two groups in the 2 RCTs (n=257) that reported this outcome. However, the inclusion of non-randomised studies (total n=1,298) resulted in a statistically significant result, favouring the intervention arm (OR 0.51, 95% CI: 0.30, 0.86).
There was limited evidence on emergence of resistance. No systematic analysis was possible and no meaningful conclusions could be drawn from the information available.
Rates of toxicity were nil or negligible in the studies that reported this outcome (3 RCTs, n=186).