BACKGROUND: Magnetic resonance imaging (MRI) is increasingly being used for rectal cancer staging. The purpose of this study was to determine the accuracy of phased array MRI for T category (T1-2 vs. T3-4), lymph node metastases, and circumferential resection margin (CRM) involvement in primary rectal cancer. METHODS: Medline, Embase, and Cochrane databases were searched using combinations of keywords relating to rectal cancer and MRI. Reference lists of included articles were also searched by hand. Inclusion criteria were: (1) original article published January 2000-March 2011, (2) use of phased array coil MRI, (3) histopathology used as reference standard, and (4) raw data available to create 2×2 contingency tables. Patients who underwent preoperative long-course radiotherapy or chemoradiotherapy were excluded. Two reviewers independently extracted data. Sensitivity, specificity, and diagnostic odds ratio were estimated for each outcome using hierarchical summary receiver-operating characteristics and bivariate random effects modeling. RESULTS: Twenty-one studies were included in the analysis. There was notable heterogeneity among studies. MRI specificity was significantly higher for CRM involvement [94%, 95% confidence interval (CI) 88-97] than for T category (75%, 95% CI 68-80) and lymph nodes (71%, 95% CI 59-81). There was no significant difference in sensitivity between the three elements as a result of wide overlapping CIs. Diagnostic odds ratio was significantly higher for CRM (56.1, 95% CI 15.3-205.8) than for lymph nodes (8.3, 95% CI 4.6-14.7) but did not differ significantly from T category (20.4, 95% CI 11.1-37.3). CONCLUSIONS: MRI has good accuracy for both CRM and T category and should be considered for preoperative rectal cancer staging. In contrast, lymph node assessment is poor on MRI.