It is suggested that the internal thoracic artery (ITA) harvesting technique influences the incidence of sternal wound infection (SWI) after coronary artery bypass graft (CABG). To determine if there is any real difference between skeletonized vs pedicled ITA, we performed a meta-analysis to determine if there is any real difference between these two established techniques in terms of SWI. We performed a systematic review using MEDLINE, EMBASE, CENTRAL/CCTR, SciELO, LILACS, Google Scholar and reference lists of relevant articles to search for studies that compared the incidence of SWI after CABG between skeletonized vs pedicled ITA until June 2012. The principal summary measures were odds ratio (OR) with 95% confidence interval (CI) and P values (statistically significant when <0.05). The ORs were combined across studies using the weighted DerSimonian–Laird random effects model and weighted Mantel–Haenszel fixed effects. Meta-analysis, sensitivity analysis and meta-regression were completed using the software Comprehensive Meta-Analysis version 2 (Biostat, Inc., Englewood, NJ, USA). Twenty-two studies involving 4817 patients (2424 skeletonized; 2393 pedicled) met the eligibility criteria. There was no evidence for important heterogeneity of effects among the studies. The overall OR (95% CI) of SWI showed a statistically significant difference in favour of skeletonized ITA (fixed effect model: OR 0.443, 95% CI 0.323–0.608, P < 0.001; random effect model: OR 0.443, 95% CI 0.323–0.608, P < 0.001). In the sensitivity analysis, the difference in favour of skeletonized ITA was also observed in subgroups such as diabetic, bilateral ITA and diabetic with bilateral ITA; we also observed that there was a difference in the type of study, since non-randomized studies together demonstrated the benefit of skeletonized ITA in comparison with pedicled ITA, but the randomized studies together did not show this difference (although close to statistical significance and with the tendency to favour the skeletonized group). In meta-regression, we observed a statistically significant coefficient for SWI and proportion of diabetic patients (coefficient −0.02, 95% CI −0.03 to −0.01, P = 0.016). In conclusion, skeletonized ITA appears to reduce the incidence of postoperative SWI in comparison with pedicled ITA after CABG, with this effect being modulated by the presence of diabetes.