COGI 2023

DEVELOPING A PREDICTIVE MODEL FOR MINIMAL-MILD ENDOMETRIOSIS AS A CLINICAL SCREENING TOOL IN INFERTILE WOMEN: UTEROSACRAL TENDERNESS AS A KEY PREDICTOR

Jie Zhang 1 Jingyi Zhang 1 Jin Liu 2 Yanhong Xu 1 Peipei Zhu 1 Lei Dai 1 Li Shu 1 Jinyong Liu 1 Zhen Hou 1 Feiyang Diao 1 Jiayin Liu 1 Jing Wang* 1 Yundong Mao* 1
1State Key Laboratory of Reproductive Medicine, Clinical Center of Reproductive Medicine, The First Affiliated Hospital of Nanjing Medical University, Nanjing
2Clinical Research Institute, The first affiliated Hospital of Nanjing Medical University, Nanjing

Problem statement: Approximately one-third of women with minimal-mild endometriosis (EMs) have infertility, and the diagnosis of these women has been long delayed. A predictive model for minimal-mild EMs with good accuracy for screening these women in primary care would reduce the time to diagnosis, however, there is lack of such predictive model due to traditional non-invasive methods are of little use for this type of EMs.

Methods: We conducted a single-center retrospective cohort study, including 1365 patients who underwent laparoscopy from January 2013 to August 2020. Patients were divided into a training set (n=910) to develop the predictive model and a validation set (n=455) to confirm the model prediction efficiency, with 2:1 ratio randomly. Preoperative history, symptoms and blood tests in the electronic medical record were collected by trained clinical researchers. Univariable and multivariable analysis were used with the training set to select independent predictors. The Hosmer-Lemeshow goodness of fit test, Net Reclassification Improvements and Integrated Discrimination Improvements were used to select the optimum model in the training set. The discriminations, calibrations and clinical use of the prediction model were tested in the both training and validation sets.

Results: Body mass index, dysmenorrhea, dyspareunia, uterosacral tenderness and serum CA-125 were the most important predictors of minimal-mild EMs. The prediction model had sensitivities of 87.7% and 93.3%, specificities of 68.6% and 66.4%, and an area under the curve of 0.84 (95% confidence interval [CI] 0.81-0.87) and 0.85 (95% CI 0.80-0.89) for the training and validation sets, respectively. Calibration curves and decision curve analysis also showed the model having good calibration and clinical value. Uterosacral tenderness was found to be the most valuable predictor.

Conclusion: This study developed a predictive model with good accuracy to identify infertile women with minimal-mild EMs based on clinical characteristics, signs and inexpensive blood test. This model would help clinicians to screen infertile women for minimal-mild EMs, facilitating the process of early diagnosis and treatment.

The nomogram (A), receiver operating characteristic curves (B), calibration plots (C) and decision curve analysis curves (D) of the prediction model.

Jie Zhang
Jie Zhang