一种预测子宫内膜病变患者内膜病理类型的临床模型＊

目的 开发并评估一种用于预测子宫内膜病变患者内膜病理类型的临床模型;方法 选取2019年11月至2021年11月因妇科B超发现子宫内膜病变并于山东中医药大学附属医院行宫腔镜下内膜活检的患者,结合其病史与最小绝对收缩和选择算子法（Least absolute shrinkage and selection operator,LASSO）筛选影响内膜病变的独立危险因素,列线图（nomogram）函数建立列线图模型,采用ROC曲线下方的面积大小（Area Under Curve, AUC）、C指数（C-index）、拟合优度（Hosmer-Lemeshow）检验、自举法（bootstrap）评估及验证模型,依据列线图对纳入患者进行风险赋分,绘制ROC曲线获取风险评分的截断值,从而划分高低风险的人群;结果 阴道流血、绝经、无流产史、并发高血压、B超内膜厚度增厚、内膜回声不正常,中医证型为虚实夹杂证是发生子宫内膜癌及癌前病变的独立危险因素,评估模型所得AUC值与C-index均>0.9,Hosmer-Lemeshow检验显示P>0.05,bootstrap内部验证法所得C-index亦高于0.9,均说明模型准确度、区分度及可信度良好,当患者的风险评分总和≥177分时,属于高风险人群;结论 本模型可以预测B超提示子宫内膜病变患者的内膜病理类型为癌前或恶变型的概率并识别高风险人群。

A Clinical Model for Predicting Pathological Types of Endometrium in Patients with Endometrial Diseases

Objective To develop and evaluate a clinical model for predicting endometrial pathological types in patients with endometrial lesions.Methods Patients who were diagnosed with endometrial lesions by gynecological B ultrasound from November 2019 to November 2021 and underwent hysteroscopic endometrial biopsy in the Affiliated Hospital of Shandong University of Traditional Chinese Medicine. Combined with their medical history and Lasso regression to screen the independent risk factors for endometrial lesions, nomogram model was established using Nomogram function. AUC value, C-index, Hosmer-Lemeshow test, and bootstrap internal validation method were used to evaluate the model. According to the nomogram, the risks of included patients were scored, and the ROC curve was drawn to obtain the cutoff value of risk score, thus dividing the high-risk and low-risk groups.Results Vaginal bleeding, menopause, no history of abortion, complicated hypertension, B ultrasound endometrial thickness thickening, abnormal endometrial echo, the TCM syndrome type of deficiency-excess syndrome was the independent risk factors of endometrial cancer and precancerous lesions. AUC value and C-index obtained from model evaluation were higher than 0.9, Hosmer-Lemeshow test showed P>0.05, and C-index obtained from bootstrap internal verification method was also higher than 0.9, all of which indicated that the model had good accuracy, differentiation degree and credibility. When the total risk score of patients was higher than or equal to177, it belonged to the high-risk group.Conclusion This model can predict the probability of B ultrasonography indicating that the pathological type of endometrium in patients with endometrial lesions is precancerous or malignant transformation and identify high-risk populations.