Clinical and pathological factors affecting the accuracy of estimating myometrial invasion on magnetic resonance imaging in early-stage endometrial cancer

Abstract

<h3>Objectives:</h3> To determine the accuracy of MRI at estimating myometrial invasion (MI) and the clinical and pathological factors that may affect this accuracy in patients with early stage endometrial cancer undergoing hysterectomy. <h3>Methods:</h3> This IRB approved, retrospective study included all patients who had an MRI within 90 days preceding hysterectomy for endometrial cancer from January 2013 through December 2019 across 3 Mount Sinai-affiliated hospitals. Charts were reviewed for clinical characteristics including BMI, menopausal status, abnormal uterine bleeding, and prior pelvic surgery. MRIs were re-reviewed by 3 radiologists blinded to original interpretations. The radiologists recorded degree of myometrial invasion (MI; <50%, or ≥50%) and presence of benign pathology findings such as fibroids, endometriosis and adenomyosis. The degree of MI from MRI was collapsed to one value representing a combination of the 3 radiologist readers. This reading was compared to the degree of MI from the final pathology specimen to calculate accuracy, sensitivity, specificity, positive predictive value, and negative predictive value of MRI evaluation. The association between patient characteristics and MRI accuracy was assessed by using the Chi-square or Fisher's exact test for categorical characteristics and the Mann-Whitney test for continuous characteristics. <h3>Results:</h3> A total of 53 patients were included; charts and MRIs were reviewed. Median patient age was 61 (range 29-86); median BMI, 29.9 (range 17.6-65.1) kg/m². On final pathology specimen, 29 patients had FIGO stage 1A (53.7%) and 10 patients with FIGO stage 1B (18.5%). The median depth of MI on pathology was 0.5cm (range 0-10cm). Degree of MI was dichotomized into ≥50% vs <50% with MI on final pathology as the gold standard. The accuracy of MRI in evaluating MI across the 3 radiologists ranged from 70-80%; sensitivity of 65-80%; specificity of 65-82%; PPV 57-70%; and NPV of 79-87%. Twenty-nine patients on final pathology were classified as stage 1A and 10 patients were classified as stage 1B. For stage 1A on final pathology, MI was accurately estimated on MRI for 23/28 (82.1%) and overestimated for 5/28 (17.9%). For stage 1B on final pathology, MI was accurately estimated on MRI for 7/10 patients (70.0%) and underestimated for 3/10 (30.0%). Patient characteristics such as BMI, prior pelvic surgery, menopausal status, and presence of abnormal bleeding did not have a statistically significant effect on the accuracy or MI estimation on MRI (p≥0.13). Benign MRI findings including fibroids, adenomyosis, and endometriosis were not found to affect accuracy of MI invasion. <h3>Conclusions:</h3> Pelvic MRI is the best radiologic modality available that is currently utilized in assessment of the depth of myometrial invasion in patients with endometrial cancer. Accurate prediction of presence or absence of invasion has significant implication on clinical management particularly in patients with early stage endometrial cancer desiring fertility preservation. In our study, presence of myometrial invasion was underestimated in 30% of patients. Patients have to be appropriately counseled with regards to MRI accuracy in predicting invasion. Neither patients' clinical characteristics nor presence of other pathological findings on MRI were found to impact the accuracy of MRI estimation of MI.