Diagnostic performance of MR imaging in evaluating prognostic factors in patients with cervical cancer: a meta-analysis

This study aims to determine the diagnostic performance of conventional magnetic resonance imaging (MRI) in assessing the distance between the tumor and the internal os, stromal infiltration, lymph node metastasis, and parametrial invasion in patients with cervical cancer. A systematic English-language literature search of conventional MRI in the evaluation of human cervical cancer was performed in the PubMed, Cochrane Library, Embase, and Web of Science databases from 1995 to 2018. The pooled sensitivity, specificity, diagnostic odds ratio (DOR), and positive and negative likelihood ratios (PLR and NLR) of all studies were calculated. The results were then plotted in a hierarchical summary receiver operating characteristic (HSROC) plot, and meta-regression and subgroup analyses of the parametrial invasion were also performed. The pooled sensitivity, specificity, DOR, PLR, and NLR were 86%, 97%, 167.91, 24.74, and 0.15, respectively, in evaluating the internal os involvement (6 studies, 454 patients); 87%, 91%, 73.41, 10.22, and 0.14, respectively, in evaluating the stromal infiltration (11 studies, 672 patients); 51%, 89%, 8.63, 4.72, and 0.55, respectively, in evaluating the lymph node metastasis (15 studies, 997 patients); and 75%, 92%, 34.01, 9.38, and 0.28, respectively, in evaluating the parametrial invasion (19 studies, 1748 patients). The meta-regression of the parametrial invasion showed that the application of contrast enhancement was a significant factor affected the heterogeneity (p = 0.039). Conventional MRI can accurately evaluate the distance between the tumor and the internal os, as well as stromal infiltration, and performs well in diagnosing the parametrial invasion. However, this method exhibited a limited ability in diagnosing the lymph node metastasis.