中国全科医学 ›› 2024, Vol. 27 ›› Issue (20): 2483-2490.DOI: 10.12114/j.issn.1007-9572.2023.0862

• 论著 • 上一篇    下一篇

前列腺癌早期筛查风险预测模型的建立与验证研究

李宏基1, 赵晓龙1, 胡伟1, 韩东晖1, 王安辉2,*(), 秦卫军1,*()   

  1. 1.710032 陕西省西安市,空军军医大学第一附属医院泌尿外科
    2.710032 陕西省西安市,空军军医大学军事预防医学系流行病学教研室
  • 收稿日期:2023-11-10 修回日期:2024-02-20 出版日期:2024-07-15 发布日期:2024-04-08
  • 通讯作者: 王安辉, 秦卫军

  • 作者贡献:

    李宏基负责临床数据收集、整理、分析,并撰写论文初稿;赵晓龙负责统计分析、绘制图表质量把控;胡伟负责临床资料质量把控;韩东晖提出研究思路,设计研究方案;王安辉、秦卫军负责文章的质量控制及审校,对文章整体负责。

  • 基金资助:
    国家自然科学基金资助重点国际(地区)合作与交流项目(822201080041005120); 国家自然科学基金资助项目(82202933); 陕西省自然科学基金资助项目(2022JQ-774)

Development and Validation of a Prediction Model for Prostate Cancer Early Screening

LI Hongji1, ZHAO Xiaolong1, HU Wei1, HAN Donghui1, WANG Anhui2,*(), QIN Weijun1,*()   

  1. 1. Department of Urology, the First Affiliated Hospital of Air Force Medical University, Xi'an 710032, China
    2. Department of Epidemiology, School of Preventive Medicine, Air Force Medical University, Xi'an 710032, China
  • Received:2023-11-10 Revised:2024-02-20 Published:2024-07-15 Online:2024-04-08
  • Contact: WANG Anhui, QIN Weijun

摘要: 背景 前列腺癌作为常见的恶性肿瘤,威胁中老年男性生命健康,其预后与初诊时肿瘤级别与分期密切相关。前列腺特异性抗原(PSA)是前列腺癌早期筛查的重要分子,但非肿瘤负荷的前列腺良性疾病或操作也会引起PSA升高,盲目穿刺常导致过度诊疗。依据患者穿刺活检前的临床指标构建前列腺癌风险预测模型,可以为早筛可疑患者是否进行穿刺活检提供重要参考。 目的 寻找前列腺穿刺活检阳性的独立危险因素,构建前列腺癌发生的风险预测模型,预测前列腺癌发生风险。 方法 选取2011年1月—2023年6月于空军军医大学第一附属医院泌尿外科住院并进行前列腺穿刺活检的患者1 138例,排除351例临床数据不完整病例后,剩余787例病例通过R语言split函数随机分为训练集(n=548)与验证集(n=239)(划分比例为7∶3)。收集患者的基本信息以及穿刺活检前生化检查指标,依据患者穿刺活检病理结果判断是否发生前列腺癌。应用LASSO回归筛选前列腺癌发生的独立危险因素,对独立危险因素进行多因素Logistic回归分析,利用分析结果构建前列腺癌早期筛查的风险预测模型并绘制列线图。依据训练集与验证集数据对模型进行验证。 结果 使用LASSO回归分析筛选出6个预测变量,包括年龄、总PSA(tPSA)、碱性磷酸酶、总蛋白、钙、尿素。多因素Logistic回归分析结果显示,年龄≥60岁(OR=3.769,95%CI=2.393~5.937)、tPSA≥10 μg/L(OR=2.259,95%CI=1.419~3.596)、碱性磷酸酶≥45 U/L[45~<125 U/L:OR=20.136,95%CI=4.419~91.752;≥125 U/L:OR=45.691,95%CI=9.199~226.951]是前列腺癌的危险因素(P<0.05),总蛋白≥65 g/L(OR=0.086,95%CI=0.031~0.236)、钙≥2.11 mmol/L(OR=0.148,95%CI=0.054~0.403)、尿素≥9.5 mmol/L(OR=0.069,95%CI=0.019~0.252)是前列腺癌的保护因素(P<0.05)。依据有统计学差异的6个预测变量绘制列线图,建立预测模型。预测模型预测训练集前列腺癌发生的ROC曲线下面积(AUC)为0.778(95%CI=0.740~0.816),灵敏度为53.2%,特异度为85.5%,验证集前列腺癌发生的AUC为0.770(95%CI=0.708~0.832),灵敏度为61.2%,特异度为80.0%。拟合优度检验显示训练集P=0.543,验证集P=0.372,具有较好的拟合度。决策曲线分析(DCA)显示训练集高风险阈值<10%,验证集的高风险阈值约为15%,在临床实践中有一定的指导意义。 结论 本研究建立了包含年龄、tPSA、碱性磷酸酶、总蛋白、钙、尿素共6项穿刺前指标的前列腺癌列线图风险预测模型,可用于预测早期筛查可疑患者的前列腺癌发生风险。

关键词: 前列腺肿瘤, 前列腺增生, 筛查, 危险因素, 列线图

Abstract:

Background

As a common malignant tumor, prostate cancer (PCa) poses a significant threat to the well-being of men worldwide. The prognosis of PCa is intricately linked to the grade and stage of the tumor at the time of initial detection. Prostate specific antigen (PSA) is a key biomarker for evaluating prostate health, yet lacks specificity for prostate cancer tumors. Elevated PSA levels can also be caused by benign prostate diseases. And the indiscriminate use of biopsy resulting in overdiagnosis. Hence, the development of a prostate cancer risk prediction model based on pre-biopsy clinical indicators in patients can serve as a valuable tool for early screening of individuals with suspicious findings warranting biopsy.

Objective

To examine the individual risk factors associated with positive prostate biopsy outcomes and develop a risk assessment model for predicting positive biopsy results in PCa screening.

Methods

A total of 1 138 patients who underwent prostate biopsy in the Department of Urology, the First Affiliated Hospital of Air Force Medical University from January 2011 to June 2023 were gathered and organized. Following the exclusion of 351 cases with inadequate clinical data, the remaining 787 cases were randomly allocated into a training set and validation set in a 7∶3 ratio by R software. Patient demographics and routine biochemical test results prior to biopsy were compiled, with PCa diagnosis determined based on the outcomes of the biopsy. LASSO regression analysis in the R software was utilized to identify independent risk factors associated with the development of PCa based on biochemical indicators. Subsequently, multivariate logistic regression analysis in SPSS software was employed to construct an early screening and predictive model for PCa, with a Nomogram being generated. The model was validated according to the data of training set and validation set.

Results

The study utilized LASSO regression analysis to identify 6 independent risk factors associated with positive prostate biopsy results, including age, total PSA (tPSA), alkaline phosphatase, serum protein level, Ca2+, and urea. Multivariate Logistic regression analysis revealed that individuals aged 60 years or older (OR=3.769, 95%CI=2.393-5.937), with tPSA levels of 10 μg/L or higher (OR=2.259, 95%CI=1.419-3.596), and alkaline phosphatase levels exceeding 45 U/L (45-<125 U/L, OR=20.136, 95%CI=4.419-91.752; ≥125 U/L, OR=45.691, 95%CI=9.199-226.951) were at increased risk for positive prostate biopsy outcomes (P<0.05). Conversely, higher levels of serum total protein (≥65 g/L, OR=0.086, 95%CI=0.031-0.236), Ca2+ (≥2.11 mmol/L, OR=0.148, 95%CI=0.054-0.403), and urea (≥9.5 mmol/L, OR=0.069, 95%CI=0.019-0.252) were found to be protective factors against positive prostate biopsy results (P<0.05). Based on the identification of 6 independent risk factors exhibiting statistically significant differences, a nomogram was constructed and a predictive model was developed. The predictive model yielded an Area under the receiver operating characteristic (ROC) curve (AUC) of 0.778 (95%CI=0.740-0.816) for PCa in the training set, with a sensitivity of 53.2% and a specificity of 85.5%. In the validation cohort, the AUC for PCa was 0.770 (95%CI=0.708-0.832), with a sensitivity of 61.2% and a specificity of 80.0%. The goodness of fit test indicated P=0.543 in the training set and P=0.372 in the validation set, demonstrating a satisfactory level of fit. The discriminant analysis (DCA) demonstrated that the high-risk threshold in the training set was below 10%, while in the validation set it was approximately 15%, indicating valuable implications for clinical practice.

Conclusion

This study developed a PCa nomogram risk prediction model incorporating 6 biochemical indicators, namely age, tPSA, alkaline phosphatase, serum total protein, Ca2+, and urea, prior to prostate biopsy, to effectively forecast PCa risk in patients with favorable early screening outcomes.

Key words: Prostatic neoplasms, Prostatic hyperplasia, Screening, Risk factors, Nomogram