Magnetic Resonance Imaging as a Valuable Tool for Predicting Parametrial Invasion in Stage IB1 to IIA2 Cervical Cancer

Objectives The aims of study were to determine the predictive value of preoperative magnetic resonance imaging (MRI) for parametrial invasion (PMI) and to develop a predictive model for PMI in patients with stage IB1 to IIA2 cervical cancer. Methods We retrospectively analyzed patients with stage IB1 to IIA2 cervical cancer (n = 215) who underwent radical hysterectomy between 2003 and 2014. The presence of PMI from postoperative pathological reports and its association with preoperative MRI findings were evaluated. We developed a predictive model for PMI using independent predictive factors identified by logistic regression and estimated its predictive performance. Results Thirty patients (14.0%) had PMI from postoperative pathological reports. Among the preoperative MRI findings, a greater tumor diameter (4.2 vs 2.0 cm; P < 0.001), a larger tumor volume (92.6 vs 12.7 cm 3 ; P < 0.001), the presence of PMI (53.3% vs 8.6%; P < 0.001), and upper vaginal involvement (73.3% vs 22.7%; P < 0.001) were significantly associated with PMI. Multivariate analysis identified tumor volume (odds ratio, 7.0; 95% confidence interval, 2.63–18.53; P < 0.001) and PMI (odds ratio, 6.1; 95% confidence interval, 2.31–15.97; P < 0.001) from preoperative MRI findings as independent predictive factors for PMI. Our predictive model demonstrates that the presence of PMI or a tumor volume of greater than 18.0 cm 3 has a higher sensitivity (86.7% vs 53.3%) and lower specificity (74.6% vs 91.4%) than the presence of PMI alone. Specifically, the model's negative predictive value was superior to that of PMI only (97.2% vs 92.3%). In the low-risk group, defined as preoperative MRI findings suggesting no PMI and a tumor volume of 18.0 cm 3 or less, the proportion of false negative cases was just 2.8%. Conclusions When tumor volume with findings suggesting that PMI is considered, preoperative MRI is useful in excluding PMI. A predictive model based on preoperative MRI findings seems to be valuable in identifying potential candidates for less radical surgery in stage IB1 to IIA2 cervical cancer.