Abstract
BACKGROUND: This retrospective study evaluated the clinical features of perineal endometriosis (PEM) and established a prognostic nomogram for recurrence probability in patients treated with surgical resection. METHODS: This study enrolled 130 PEM patients who had received surgical treatment in Peking Union Medical College Hospital (PUMCH) between January 1992 and September 2020. We collected their clinical features and conducted outpatient or telephone follow-up. The predictive nomogram was constructed based on 104 patients who had completed follow-up by July 2021. The Cox proportional hazards regression model was used to evaluate the prognostic effects of multiple clinical parameters on recurrence. The Index of concordance (C-index) and calibration curves were used to access the discrimination ability and predictive accuracy of the nomogram respectively, and the results were further validated via bootstrap resampling. Calculating the area under the curve (AUC) via risk scores of patients aimed to further access the predictive power of the model. In addition, the survival curve was depicted using Kaplan-Meier plot and compared by log-rank method. RESULTS: Most PEM patients had been symptomatic for 24-48 months before the lesion resection. With a median 99.00 (interquartile range: 47.25-137.50) months of postoperative observation, there were 16 (15.1%) out of 104 cases who finished follow-up reported symptomatic recurrence. On multivariate analysis of derivation cohort, multiple lesions, microscopically positive margin (mPM) and anal sphincter involvement (ASI) were selected into the nomogram. The C-index of the nomogram for predicting recurrence was 0.84 (95% CI 0.77-0.91). The calibration curve for probability of recurrence for 36, 60 and 120 months showed great agreement between prediction by nomogram and actual observation. Furthermore, the AUCs of risk score for 36, 60 and 120 months were 0.89, 0.87 and 0.82 respectively. CONCLUSIONS: PEM is a rare kind of endometriosis and surgery is the primary treatment. Multiple lesions and ASI are independent risk factors for postoperative recurrence, and wide resection with more peripheral tissue could be preferred. The proposed nomogram resulted in effective prognostic prediction for PEM patients receiving surgical excision. In addition, this predictive nomogram needs external data sets to further validate its prognostic accuracy in the future.