A variety of studies have been performed to elucidate the polymorphisms in promoter regions of matrix metalloproteinases (MMPs) associated with the risk of digestive cancers, and yet, results remain conflicting and heterogeneous. Thus, we undertook a systematic meta-analysis to determine the genetic susceptibility of MMPs to digestive cancers. A computerized literature search was conducted in databases of PubMed, Embase, and ISI Web of Knowledge till October 2012 for any MMP genetic association study in oral squamous, gastric, esophageal, and colorectal carcinomas. Odds ratios (OR) and 95% confidence interval (CI) were estimated for each gene under dominant and recessive models, and the heterogeneity between studies was assessed using Q test and I (2) value. Overall and subgroup analysis according to anatomical sites and ethnicity was carried out. Statistical analysis was performed with Review Manager 5.0. A total of 40 eligible publications with 68 comparisons were included in this study. For MMP1 nt-1607, individuals with 2G state could increase risk of digestive cancers in total analysis (dominant: OR=1.31, 95% CI=1.16-1.48, P<0.00001; recessive: OR=1.29, 95% CI=1.11-1.50, P=0.0009). In the subgroup of tumor sites, significant associations were also observed in esophageal cancer and colorectal cancer under both genetic models. For MMP2 nt-1306, CT or TT carriers performed significant protection against digestive cancer in the dominant model (OR=0.69, 95% CI=0.55-0.85, P=0.0007) of the overall. In the subgroup analysis, significant association was found in esophageal cancer, with borderline effects in gastric cancer and oral squamous cell carcinoma. For MMP7 -181 A/G, significant association was observed under two genetic models in the overall (dominant: OR=1.26, 95% CI=1.10-1.43, P=0.0009; recessive: OR=1.33, 95% CI=1.11-1.60, P=0.002) and in the individual cancer subgroup of esophageal cancer and gastric cancer. For MMP9 -1,562 C/T, a borderline effect was found with digestive cancers in the total and stratified analysis of the colorectal cancer under dominant model. No association was observed in either the overall or subgroup analysis for MMP3 -1,171 5A/6A. Our meta-analysis demonstrated the fact that polymorphisms in promoter regions of MMP genes might be related to the susceptibility of digestive cancers, with cancer development for MMP1 and MMP7, and a protection against cancer for MMP2 and MMP9. Further evidences with adequate sample sizes need to be conducted.