摘要
Abstract
Objective:To construct and validate a nomogram model of endometrial cancer(EC)p53 aberrant type(p53abn)molecular subtypes.Methods:433 EC patients ad-mitted to Qilu Hospital of Shandong University from January 2021 to December 2024 were retro-spectively included,divided into p53abn and non-p53abn groups(including POLEmut/MMRd/NSMP subtypes)based on the results of molecular typing assay,and classified into a training set(n=284)and a validation set(n=149)based on the time cut-off point.Independent predictors were screened by multifactor logistic regression,and a column-line graphical model was devel-oped and evaluated for performance:subject work characteristic curve(ROC)was used to ana-lyze discrimination,calibration curve to assess calibration,decision curve analysis(DCA)to quantify clinical utility,and internal validation of the model.Results:Multifactorial logistic re-gression analysis of the training set showed that age≥56 years(OR=11.23,95%CI:4.029～31.306),NLR≥2.78(OR=16.743,95%CI:6.286～44.594),PROGRP≥45.19(OR=6.503,95%CI:2.507～16.864),Ferr≥84.26(OR=3.363,95%CI:1.324～8.546),and the degree of differentiation(moderately differentiated OR=0.091,95%CI:0.028～0.289;hy-perdifferentiated OR=0.039,95%CI:0.01～0.147)were the independent p53abn subtype predictors(all P<0.001).The column-line graph model had an AUC of 0.964 in both the training and validation sets,and the calibration curve showed that the predicted risk was highly consistent with the actual risk,with good calibration ability and net clinical benefit.Conclu-sion:The nomogram model constructed in this study may provide a reliable tool for individual-ized prediction of p53abn molecular subtypes in endometrial cancer and clinical decision-mak-ing.关键词
机器学习/子宫内膜癌/分子分型/p53abn/列线图Key words
Machine learning/Endometrial cancer/Molecular typing/p53abn/Nomo-gram model分类
医药卫生
