Standard 1.5-T MRI of endometrial carcinomas: modest agreement between radiologists

Abstract

To evaluate pelvic magnetic resonance imaging (MRI) interobserver agreement for the detection of deep myometrial invasion, cervical stroma invasion and lymph node metastases in endometrial carcinoma patients in relation to surgical staging. Fifty-seven patients with histologically confirmed endometrial carcinoma were prospectively included in a study of preoperative 1.5-T MRI. Four radiologists, blinded to patient data, independently reviewed the images for the presence of deep myometrial invasion, cervical stroma invasion and lymph node metastases. Kappa coefficients for interobserver agreement and diagnostic performances for each observer were calculated using final surgical staging results (FIGO 09) as reference standard. Overall agreement among all observers was moderate for cervical stroma invasion (κ = 0.50 [95% CI 0.27-0.73]) and lymph node metastases (κ = 0.56 [0.09-0.80]) and fair for deep myometrial invasion (κ = 0.39 [0.26-0.55]). Sensitivity (specificity) values for the four observers were 72-92% (44-63%) for deep myometrial invasion, 38-63% (82-94%) for cervical stroma invasion and 25-38% (90-100%) for lymph node metastases. Conventional MRI showed only modest interobserver agreement and diagnostic accuracy for detection of deep myometrial invasion, cervical stroma invasion and lymph node metastases. Improved methods are needed for preoperative imaging in the staging of endometrial carcinomas. MRI is an important tool for preoperative endometrial cancer staging. • Staging agreement based on pelvic MRI was modest among different observers. • Preoperative MRI alone was suboptimal in identifying high-risk patients. • Improved imaging and biomarkers may refine preoperative risk stratification in endometrial cancer.