An Ultrasound-based Prediction Model to Predict Ureterolysis during Laparoscopic Endometriosis Surgery

Abstract

Study objective: To develop a model, including clinical features and ultrasound findings, to predict the need for ureterolysis (i.e., dissection of the ureter) during laparoscopy for endometriosis.

Design: A retrospective observational study of patients who had undergone transvaginal ultrasound (TVS) according to the International Deep Endometriosis Analysis consensus and subsequent laparoscopy ± excision of endometriosis between January 2017 and February 2021 was conducted.

Setting: Sydney Medical School Nepean, University of Sydney, Nepean Hospital, and Blue Mountains Hospital, New South Wales, Australia.

Patients/participant: 177 patients.

Intervention: The demographic, clinical, TVS, and intraoperative data were extracted through electronic clinical records.

Measurements and main results: Multicategorical decision-tree and baseline models were built to choose the variables most correlated to the outcome under study. Receiver operating characteristic analysis was performed on the binary classification. Based on our results, we selected the variables performing with significant statistical differences (p <.05). During the study period, 177 consecutive patients were recruited and divided into 2 subgroups, ureterolysis (51.4%) and nonureterolysis (48.6%). Ureterolysis was noted in 87.5% of patients in which the left ovary was immobile (p <.001) and in 82.5% in which the right ovary was fixed (p <.001). For patients with right uterosacral ligament (USL) deep endometriosis (DE), ureterolysis was performed in 96.2% patients (p <.001) and 64.6% (p = .043) for left USL DE. Among patients with bowel DE, the proportion of patients undergoing ureterolysis was 95.5% (p <.001). The prognostic variables used in the final model to predict ureterolysis included dyschezia, absence of ovarian mobility, presence of right or left USL DE, and presence of bowel DE on TVS. According to the developed model, the baseline risk for performing ureterolysis is 20% in our sample. The overall model performance demonstrated an area under the receiver operating characteristic curve 0.82.

Conclusion: Our study demonstrates that it is possible to predict the need for ureterolysis with clinical and sonographic data. Furthermore, patients presenting with a combination of the variables of our model (dyschezia, ovarian immobility, USL, and bowel DE lesions) have a high risk of ureterolysis. In contrast, patients without these features have a low risk (approximately 20%) of needing ureterolysis.

Keywords: Endometriosis surgery; Laparoscopy; Prediction model; Ultrasound; Ureterolysis.