Click on image to view larger version.

Figure 4 Funnel-plot analysis. Plot of the logarithm of the relative risk of mortality versus precision of the included trials. The study by Marini et al. (7) was excluded because the relative risk could not be calculated, as no deaths occurred in either treatment group. The funnel plot is symmetrical (intercept 0.15; 95% confidence interval –1.06 to 1.37), contradicting the presence of a publication bias.