Objective The aim of this study was to review the available literature on the reliability of contemporary magnetic resonance imaging (MRI) techniques in the assessment of high-risk features of endometrial carcinoma, that is, deep myometrial invasion, cervical stromal involvement, and lymph node metastasis.

Methods The PubMed and Scopus databases were searched for studies published before March 2014. Studies on plain MRI were excluded.

Results Fifty-two eligible studies were identified. For the assessment of deep (≥50%) myometrial invasion (50 studies, 3720 patients), the pooled sensitivity, specificity, positive predictive value, and negative predictive value were 80.7%, 88.5%, 77.6%, and 89.5%, respectively, by random-effects analysis. For predicting cervical stromal involvement (12 studies, 1153 patients), the pooled values were 57.0%, 94.8%, 68.7%, and 90.5%, respectively. For lymph node metastasis on a per-patient basis (10 studies, 862 patients), they were 43.5%, 95.9%, 66.3%, and 92.2%, respectively. In a meta-regression analysis, dynamic imaging was associated with a higher sensitivity in detecting deep myometrial invasion, as compared with contrast-enhanced imaging (P = 0.021). The improvement by diffusion-weighted imaging was of a borderline significance (P = 0.057). Significant small-study effects were found for the sensitivity of MRI in detecting deep myometrial invasion (P < 0.0001) and cervical stromal involvement (P = 0.049).

Conclusions Considering the poor-to-moderate sensitivity of MRI in detecting high-risk features of endometrial carcinoma, patients with negative findings on MRI may not safely forgo surgical staging unless the findings are confirmed by a backup method. The high specificities allow the targeting of staging procedures by MRI alone in patients with positive findings. Compared with contrast-enhanced imaging, dynamic and diffusion-weighted imaging may be more reliable in the radiological staging of endometrial carcinoma.