Evaluating Inter-Reader Reliability When Using MRI in Deeply Infiltrative Endometriosis

INTRODUCTION: Endometriosis is a common gynecologic disorder that, on average, affects 6–10% of women. Despite the high prevalence of the disease, its diagnosis can be complicated and often takes approximately 7–11 years to diagnose. The gold standard of diagnosis is surgery; however, this can be associated with a large economic burden. Therefore, imaging has been increasingly utilized to evaluate disease burden. While ultrasound (US) is a more accessible modality, magnetic resonance imaging (MRI) has been increasingly used to aid with preoperative planning. Several different protocols have been suggested to better evaluate endometriosis, and while different protocols exist, there is limited information on inter-reader reliability and how radiologists' interpretation compares to intraoperative findings. OBJECTIVE: The objective of this study is to determine the inter-reader reliability among in-training radiologists on MRI evaluation of endometriosis. METHODS: We conducted a retrospective cohort study of thirty-six patients who previously underwent preoperative MRI with a specialized endometriosis protocol followed by surgical management by a single surgeon at the University of California, Los Angeles. Intraoperative surgical findings were all scored by the 2021 AAGL endometriosis classification to designate the staging of disease based on evidence of pathology in different anatomical areas. The imaging was then retrospectively reviewed by four abdominal radiology fellows who used the same classification system to determine a preoperative stage. A Kappa analysis was then performed to determine inter-reader reliability of radiology-reviewed preoperative MRIs to the gold standard of intraoperative findings. RESULTS: The data suggested that there was poor agreement between in-training radiologists for determining less advanced stages of endometriosis (stage I–III) and moderate agreement when determining severe stage (stage IV). With regards to each anatomic area, there was moderate agreement (0.4–0.6) when evaluating the ovaries, and agreement was poor (<0) to fair (0.2–0.4) with regard to other anatomic areas specified in the AAGL staging system (Figure 1). CONCLUSIONS: Our data suggest that there is a lot of room for growth in inter-reader reliability when evaluating endometriosis. Overall, there is at most a moderate inter-reader reliability with identifying stage IV endometriosis preoperatively, which is similar to what other studies have seen. However, agreement in the anatomic areas we were most interested in exploring as listed on the slide was poor to slight. We were most interested in these areas because assessing disease in these anatomic areas is key for surgical planning. It helps identify individuals who will have increased complexity at the time of surgery and may need subspeciality consultation such as colorectal surgery. Future directions include performing a similar study with a larger sample size or with attending radiologists as reviewers (Tables 1 and 2).