Objective To construct a risk prediction model for endometrial cancer(EC)in postmenopausal women based on conditional inference tree.Method A total of 114 postmenopausal women who underwent panhysterectomy were selected,55 patients with endometrioid adenocarcinoma were included in the study group,46 elderly patients with atrophic endometrium and 13 patients with proliferative endometrium were included in the control group.The clinical da-ta of all patients were recorded,and the influencing factors of postmenopausal EC were analyzed by Logistic regression model.A total of 114 patients were divided into training set(79 cases)and validation set(35 cases)according to a ratio of 7∶3.A conditional inference tree model was constructed in the training set to stratify the risk of EC in postmenopausal women.In the validation set,the predictive value of the conditional inference tree model for EC in postmenopausal wom-en was analyzed by using receiver operating characteristic(ROC)curve.Result Multivariate Logistic regression analy-sis showed that increased endometrial thickness,postmenopausal bleeding(PMB)and uterine cavity occupation were in-dependent risk factors for EC in postmenopausal women(P<0.05).According to PMB and endometrial thickness,a condi-tional inference tree was constructed and the postmenopausal women were divided into EC high,medium and low risk groups:PMB,regardless of whether the endometrial thickened belonged to the high risk group;no PMB,but the endome-trial thickness>3 mm belonged to the medium risk group;no PMB,and endometrial thickness≤3 mm belonged to the low risk group.The area under the curve of conditional inference tree model for predicting EC in postmenopausal women was 0.943,the accuracy was 94.3%,the sensitivity was 94.4%,the specificity was 94.1%,the positive predictive value was 94.44%,and the negative predictive value was 94.12%.Conclusion Risk stratification based on endometrial thickness and PMB can provide reference for the initial diagnosis of EC in postmenopausal women.
endometrial cancerpostmenopausalconditional inference tree
NETL
NSTL
万方数据
