Background: To compare the effectiveness and safety of the deltopectoral approach with the deltoid-splitting approach for proximal humeral fractures treatment. Methods: We retrieved literature published from January 2002 to July 2016 in the main medical search engines (Medline, PubMed, Embase, Cochrane Library and Wanfang Database) for RCTs and non-RCTs comparing the deltoid-splitting approach with the deltopectoral approach for proximal humeral fractures treatment. Outcomes include the Constant score, operation time, and intraoperative blood loss, as well as the rates of nerve injury, avascular necrosis and non-union. Meta-analyses were performed with Cochrane Collaboration's RevMan 5.0 software. Results: Eight articles with 635 patients were included in this meta-analysis. The deltoid-splitting approach and deltopectoral approach were used to treat 299 and 336 patients, respectively. Compared with the deltopectoral approach group, the deltoid-splitting approach group were associated with better functional scores (SMD=0.46; 95% CI 0.04 to 0.88), reduced operation time (SMD=-0.65; 95% CI -1.09 to -0.21), less intraoperative blood loss (SMD=-1.66; 95% CI -3.08 to -0.25), and a lower rate of avascular necrosis (RR=0.23; 95% CI, 0.08 to 0.64). There was no significant difference in the incidence of nerve injury (RR=4.56; 95% CI 0.53 to 39.11) and non-union (RR=0.49; 95% CI, 0.13 to 1.92). Conclusion: The deltoid-splitting approach technique shows less operative time and blood loss, a lower rate of avascular necrosis and more effective recovery of shoulder function for proximal humeral fracture treatment than does the deltopectoral approach technique. Moreover, the deltoid-splitting approach dose not increase the risk of nerve injury and non-union.