膀胱鳞状细胞癌预后预测模型的构建和验证

Construction of prognostic prediction model for bladder squamous cell carcinoma

目的:分析膀胱鳞状细胞癌预后影响因素,构建预后预测模型。方法:从SEER数据库提取2000年至2019年病理诊断为膀胱鳞状细胞癌286例的临床特征及随访资料构成训练集,用于构建模型。同时收集2012年至2021年就诊于河南省中医院泌尿外科的20例膀胱鳞状细胞癌患者构成外部验证集。利用训练集数据,使用Cox回归模型筛选出膀胱鳞状细胞癌预后预测因子,基于回归结果绘制列线图。利用C指数、ROC曲线、校准曲线和决策曲线分析(DCA)评估模型的预测价值。结果:Cox回归结果显示年龄≥70岁、N1~N3期、M1期和接受非根治术的膀胱鳞状细胞癌患者预后差,HR(95%CI)分别为2.208(1.629~2.992)、2.683(1.684~4.274)、3.155(1.880~5.293)、3.288(0.805~13.432)、3.073(1.719~5.494)和2.218(1.678~2.931)。利用上述变量及回归模型,绘制1、3和5 a生存率预测的列线图。在训练集和外部验证集中,模型的C指数(95%CI)分别为0.739(0.703~0.776)和0.759(0.651~0.868)。在训练集中,1、3和5 a生存率预测的ROC曲线下面积(95%CI)分别为0.779(0.725~0.834)、0.750(0.695~0.806)和0.741(0.682~0.799),在外部验证集中分别为0.748(0.570~0.926)、0.679(0.534~0.823)和0.650(0.523~0.778)。在训练集和外部验证集,校准曲线与理想曲线接近,同时DCA曲线远离基准线,表明预测结果与实际结局具有较好的一致性。结论:构建的膀胱鳞状细胞癌预后预测模型具有较好的预测价值。

Aim:To analyze the prognostic factors of bladder squamous cell carcinoma(BSCC)and develop a prognostic prediction model.Methods:Clinical characteristics and follow-up data of 286 BSCC cases confirmed by pathology between 2000 to 2019 obtained from the SEER database were collected(traning set).Additionally,an external validation set comprising 20 patients was obtained from the Urology Department of Henan Provincial Hospital of TCM from 2012 to 2021.Cox regression model was used to identify prognostic predictors,and a nomogram was created based on the regression results in training set.The performance of the model was assessed using C-index,ROC curve,calibration curve,and decision curve analysis(DCA).Results:Cox regression analysis revealed that age ≥70 years,N1-N3 stage,M1 stage and non-radical cystectomy were associated with poor prognosis,with HR(95%CI) of 2.208(1.629-2.992),2.683(1.684-4.274),3.155(1.880-5.293),3.288(0.805-13.432),3.073(1.719-5.494),and 2.218(1.678-2.931),respectively.The model′s performance was evaluated both in the training set and external validation set,with a C-index(95%CI) of 0.739(0.703-0.776) and 0.759(0.651-0.868),respectively.The areas under the ROC curve(95%CI) for predicting survival rates at 1,3,and 5 a in the training set were 0.779(0.725-0.834),0.750(0.695-0.806),and 0.741(0.682-0.799),respectively;in the external validation set were 0.748(0.570-0.926),0.679(0.534-0.823),and 0.650(0.523-0.778).The calibration curve closely resembled the ideal curve,and the DCA decision curve deviated significantly from the baseline,indicating good agreement between predicted and actual outcomes.Conclusion:The developed prognosis prediction model demonstrates promising predictive value for BSCC.