Four RCTs and two pseudo-RCTs (n=843) were included in the meta-analysis. A further six non-randomised studies could not be included in the primary meta-analysis.
Lower wound infection rates for antibiotic-treated patients were reported in all six trials. However, the results were not statistically significant for any individual study. Combining the studies in a random-effects model demonstrated a significant effect in favour of antibiotic prophylaxis; the pooled odds ratio (OR) was 0.37 (95% confidence interval, CI: 0.17, 0.78, p<0.01). There was no evidence of statistical heterogeneity (p>0.98). The use of risk differences gave a difference of 2.9% in favour of antibiotic therapy, giving a number-needed-to-treat of 34 patients to prevent one infection. The results from meta-analyses using a fixed-effect model and Bayesian methods gave essentially identical results.
The results of the fixed-effect cumulative meta-analysis showed that the pooled OR became significant at the p<0.05 level in 1990 and at the p<0.01 level in 1994, the date of the last trial to be published.
The author states that the results of the funnel plot and Rosenthal's test indicated that there was no evidence of publication bias.
The sensitivity analyses showed that the exclusion of pseudo-randomised trials from the meta-analysis produced essentially identical results to the main analysis.
There was no evidence of different treatment effects with different dosing periods or with the inclusion of Gram-negative coverage (i.e. antibiotics with both Gram-negative and -positive coverage as opposed to Gram-positive coverage alone). There was considerable heterogeneity in the infection rates, which suggested differences in baseline infection risks between the trials. This finding was investigated using a Bayesian meta-regression. The model indicated no evidence for a dependence of the treatment effect on the control arm infection rate.
Non-randomised trials (6 studies, n=6,930): the pooled OR showed significantly lower infection rates in the prophylactic antibiotic therapy group (OR 0.22, 95% CI: 0.15, 0.33, p<0.001). There was no significant evidence of heterogeneity (p=0.76). The results of the pooled OR were dominated by a single large trial. The author states that the funnel plot showed evidence of publication bias.