Five polymorphisms, rs2293152, rs4796793, rs12949918, rs6503695, rs744166, in the STAT3 gene have been implicated in susceptibility to cancer, but the results were inconclusive. The aim of this meta-analysis is to investigate the association between the five polymorphisms and cancer risk. All eligible case-control studies published up to March 2015 were identified by searching PubMed, Web of Science, Wanfang, VIP, and CNKI. Effect sizes of odds ratio (OR) and 95 % confidence interval (95 % CI) were calculated by using a fixed- or random-effect model. A total of 15 articles were included. Overall, a significantly decreased risk was found for rs12949918 polymorphism (dominant model: OR = 0.83, 95 % CI: 0.75-0.91, recessive model: OR = 0.77, 95 % CI: 0.68-0.87, TC vs. TT: OR = 0.87, 95 % CI: 0.79-0.96, CC vs. TT: OR = 0.71, 95 % CI: 0.62-0.81), and for rs744166 polymorphism (recessive model: OR = 0.75, 95 % CI: 0.58-0.98; GG vs. AA: OR = 0.68, 95 % CI: 0.51-0.90), while there was no significant association for other three polymorphisms under all genetic models. In subgroup analysis by ethnicity, for rs12949918 polymorphism, similar results were detected among Caucasians, similarly, a significant decreased risk was observed in Asians under dominant and CC vs. TT model; for rs2293152 polymorphism, significant association was detected among Asians under recessive model. This meta-analysis suggests that the STAT3 rs12949918 and rs744166 polymorphisms, but not other three polymorphisms, may be an important protective factor for cancer.