Meta-analyses of randomised controlled trials of selective digestive decontamination have clinical outcome measures, mainly pneumonia and mortality. This meta-analysis has a microbiological endpoint and explores the impact of selective digestive decontamination on Gram-negative and Gram-positive carriage and severe infections. We searched electronic databases, Cochrane Register of Controlled Trials, previous meta-analyses and conference proceedings with no language restrictions. We included randomised controlled trials which compared the selective digestive decontamination protocol with no treatment or placebo. Three reviewers independently applied selection criteria, performed the quality assessment and extracted the data. The outcome measures were carriage and severe infection due to Gram-negative and Gram-positive bacteria. Odds ratios were pooled with the random effect model. Fifty-four randomised controlled trials comprising 9473 patients were included; 4672 patients received selective digestive decontamination and 4801 were controls. Selective digestive decontamination significantly reduced oropharyngeal carriage (odds ratio [OR] 0.13, 95% confidence interval [CI] 0.07 to 0.23), rectal carriage (OR 0.15, 95% CI 0.07 to 0.31), overall infection (OR 0.17, 95% CI 0.10 to 0.28), lower respiratory tract infection (OR 0.11, 95% CI 0.06 to 0.20) and bloodstream infection (OR 0.35, 95% CI 0.21 to 0.67) due to Gram-negative bacteria. Reduction in Gram-positive carriage was not significant. Gram-positive lower airway infections were significantly reduced (OR 0.52, 95% CI 0.34 to 0.78). Gram-positive bloodstream infections were not significantly increased (OR 1.03, 95% CI 0.75 to 1.41). The association of parenteral and enteral antimicrobials was superior to enteral antimicrobials in reducing carriage and severe infections due to Gram-negative bacteria. This meta-analysis confirms that selective digestive decontamination mainly targets Gram-negative bacteria; it does not show a significant increase in Gram-positive infection.