基于临床因素的tPSA 4~10 ng/ml前列腺癌风险分析  

Risk analysis of tPSA 4-10 ng/ml prostate cancer based on clinical factors

作　　者：王亚申 罗飞[1] 李健[1] WANG Yashen;LUO Fei;LI Jian(Department of Urology,Tianjin Union Medical Centre,Tianjin 300121,China)

机构地区：[1]天津市人民医院泌尿外科,天津300121

出　　处：《现代泌尿生殖肿瘤杂志》2025年第1期14-18,共5页Journal of Contemporary Urologic and Reproductive Oncology

摘　　要：目的通过识别肿瘤风险因素,构建前列腺癌的预测模型,为总前列腺特异抗原(tPSA)4~10 ng/ml的前列腺癌患者的临床早期筛查提供强力参考。方法回顾性收集2016年12月至2024年6月天津市人民医院泌尿外科tPSA 4~10 ng/ml的前列腺癌患者69例与前列腺增生患者132例的临床数据资料,经倾向评分匹配调整后,选取98例(每组49例)患者,比较两组患者各临床指标的组间差异,对有统计学意义的临床指标进行单因素及多因素Logistic回归分析确定前列腺癌独立危险因素并建立回归预测模型,得出预测模型计算公式,比较预测模型与肿瘤危险因素的曲线下面积(AUC),校准曲线和决策曲线分别用于评估预测模型的校准度和净获益率。结果对前列腺癌和前列腺增生两组患者一般资料进行比较,两组间患者年龄、BMI、前列腺体积、tPSA、游离前列腺特异抗原、游离/总前列腺特异抗原、前列腺特异抗原密度、乳酸脱氢酶、肌酸激酶、肌酸激酶同工酶、低密度脂蛋白胆固醇、白蛋白、Ca共13个指标差异有统计学意义,为前列腺癌临床危险因素。单因素及多因素Logistic回归分析提示前列腺体积、游离/总前列腺特异抗原、前列腺特异抗原密度、乳酸脱氢酶、低密度脂蛋白胆固醇为独立危险因素。联合独立危险因素构建前列腺癌的Logistic回归模型,该模型敏感度、特异性分别为91.8%、87.9%,阳性预测值71.2%,阴性预测值97.1%,其AUC值为0.956(P<0.001,95%CI:0.924~0.989),高于各独立危险因素的AUC值。预测模型的校准曲线结果显示模型具有良好的校准度,平均绝对误差为0.012。决策曲线显示当风险阈值大于等于0.1时,该预测模型就能获得净收益。结论基于临床因素构建的tPSA 4~10 ng/ml的前列腺癌预测模型,可为前列腺癌患者的临床早期诊断提供参考价值。Objective To construct a prediction model for prostate cancer by identifying tumor risk factors and provide strong reference for clinical early screening of prostate cancer patients with tPSA 4-10 ng/ml.Methods Clinical data of 69 cases of prostate cancer with tPSA 4-10 ng/ml and 132 prostate hyperplasia patients in the Department of Urology,Tianjin Union Medical Centre Decomber,2016 to June,2024 were retrospectively collected.After propensity score matching adjustment,98 patients(49 in each group)were selected.The intergroup differences of various clinical indicators between the two groups were compared.Univariate and multivariate logistic regression analyses were conducted on the statistically significant clinical indicators to determine the independent risk factors in prostate cancer and establish a regression prediction model.The calculation formula of the prediction model was obtained,and the area under the curve(AUC)of the prediction model and the tumor risk factors were compared.Calibration curves and decision curves were respectively used to evaluate the calibration degree and net benefit rate of predictive models.Results The general data of patients with prostate cancer and prostate hyperplasia were compared.A total of 13 indicators including age,BMI,PV,tPSA,fPSA,f/tPSA,PSAD,LDH,CK,CK-MB,LDLc,ALB and Ca were found to be statistically significant,and these were considered to be the clinical risk factors in prostate cancer.Univariate and multivariate logistic regression analyses suggested that PV,f/tPSA,PSAD,LDH,and LDLc were independent risk factors.The logistic regression model of prostate cancer was constructed by combining the independent risk factors.The sensitivity and specificity of the model were 91.8%and 87.9%,respectively,with a positive predictive value of 71.2%and a negative predictive value of 97.1%,with an AUC value of 0.956(P<0.001,95%CI:0.924-0.989),which was higher than that of the AUC value of each independent risk factor.The calibration curve results of the predictive model indicate that

关 键 词：前列腺增生 前列腺癌 前列腺特异抗原 独立危险因素 

分 类 号：R73[医药卫生—肿瘤]