基于超声特征与炎症指标列线图模型的膀胱尿路上皮癌病理分级术前无创预测研究

陶乐 张浩 周群群 林婷婷 樊丹 卢畅 黄禾菁

引用本文: 陶乐,张浩,周群群,等. 基于超声特征与炎症指标列线图模型的膀胱尿路上皮癌病理分级术前无 创预测研究[J]. 海军军医大学学报,2025,46(10): 1304-1312. DOI: 10.16781/j.CN31-2187/R.20250280.
Citation: TAO L, ZHANG H, ZHOU Q, et al. Preoperative noninvasive prediction of pathological grading of urothelial carcinoma of bladder with a nomogram model based on ultrasound features and inflammatory indicators[J]. Acad J Naval Med Univ, 2025, 46(10): 1304-1312. DOI: 10.16781/j.CN31-2187/R.20250280.

基于超声特征与炎症指标列线图模型的膀胱尿路上皮癌病理分级术前无创预测研究

doi: 10.16781/j.CN31-2187/R.20250280
基金项目: 

上海市人工智能创新发展专项 2020-RGZN-02044.

详细信息
    作者简介:

    陶乐,硕士生,住院医师.E-mail: taolele2021@163.com.

    通讯作者:

    黄禾菁, E-mail: huanghejinga@163.com.

Preoperative noninvasive prediction of pathological grading of urothelial carcinoma of bladder with a nomogram model based on ultrasound features and inflammatory indicators

Funds: 

Shanghai Artificial Intelligence Innovation Development Program 2020-RGZN-02044.

  • 摘要:  目的 探讨超声特征联合炎症指标构建的列线图模型术前无创预测膀胱尿路上皮癌(UCB)病理分级的价值。 方法 回顾性分析471例经病理确诊的UCB患者资料,其中高级别组401例、低级别组70例。收集患者一般临床资料(性别、年龄、肉眼血尿等)、超声特征(病灶部位、血流信号等)和血液炎症指标[中性粒细胞与淋巴细胞比值(NLR)等],通过单因素和多因素logistic回归分析筛选独立预测因素并构建列线图模型,采用ROC曲线、校准曲线及决策曲线分析评估模型性能。 结果 多因素logistic回归分析显示,性别(OR=2.68)、年龄(OR=1.08)、肉眼血尿(OR=3.19)、病灶位于三角区(OR=4.59)、血流信号阳性(OR=2.87)和NLR(OR=1.03)是高级别UCB的独立预测因素(均P<0.05)。多因素联合模型(一般临床特征+超声特征+炎症指标)预测高级别UCB的AUC为0.892,高于一般临床特征模型(AUC=0.799)和一般临床特征+超声特征模型(AUC=0.856);校准曲线显示多因素联合模型的预测概率与实际结果一致性良好,决策曲线分析证实其临床净获益最优。 结论 整合一般临床特征、超声特征和血液炎症指标的列线图模型可有效预测UCB病理分级,为术前无创评估提供了可靠工具,有助于指导个体化治疗决策。

     

    Abstract:  Objective To evaluate the value of a nomogram model based on ultrasonographic features and inflammatory indicators in the preoperative noninvasive prediction of pathological grading of urothelial carcinoma of bladder (UCB). Methods A retrospective analysis was conducted on 471 patients with pathologically confirmed UCB, and the patients were assigned to high-grade group (401 cases) or low-grade group (70 cases). Basic clinical data (gender, age, macroscopic hematuria), ultrasonographic features (lesion location, blood flow signal, etc.), and blood inflammatory indicators (e.g. neutrophil-to-lymphocyte ratio [NLR]) were collected. Independent predictors were screened using univariate and multivariate logistic regression, and a nomogram model was constructed. Model performance was evaluated using the receiver operating characteristic (ROC) curve, calibration curve, and decision curve analysis (DCA). Results Multivariate logistic analysis identified gender (odds ratio [OR] =2.68), age (OR=1.08), macroscopic hematuria (OR=3.19), lesion located in the trigone (OR=4.59), positive blood flow signal (OR=2.87), and NLR (OR=1.03) were independent predictors of high-grade UCB (all P < 0.05). The combined model (clinical features+ultrasonographic characteristics+inflammatory indicators) achieved an area under curve (AUC) of 0.892, which was significantly higher than the clinical feature-only model (AUC=0.799) and the clinical+ultrasonographic model (AUC=0.856). The calibration curve demonstrated good consistency between predicted and actual outcomes, and DCA confirmed its optimal clinical net benefit. Conclusion The nomogram model integrating clinical features, ultrasonographic characteristics, and inflammatory indicators can effectively discriminate UCB pathological grading, providing a reliable preoperative noninvasive assessment tool for personalized treatment decisions.

     

  • 膀胱尿路上皮癌(urothelial carcinoma of bladder,UCB)是泌尿系统最常见的恶性肿瘤之一,全球每年新发病例超过57万例,死亡病例超过20万例[1-2]。UCB具有高复发率和易进展性,严重影响患者的生活质量与长期预后[3]。肿瘤的病理分级是指导治疗决策和预后评估的关键依据。其中,高级别UCB较低级别UCB更易进展为肌层浸润性膀胱癌,且复发率更高,通常需要接受根治性膀胱切除术治疗[4]。当前,膀胱镜检查联合组织活检仍是确诊UCB病理分级的标准方法,但该方法因具有侵袭性、成本高及患者依从性差,限制了其术前的广泛应用[5]。因此,开发一种简便、无创且准确的术前预测工具以识别高级别UCB患者,对于优化治疗决策、提高治疗效果及改善患者预后具有重要临床意义。

    超声检查作为一种无创、便捷且可重复的影像学手段,在膀胱癌的初步筛查和术后随访中已被广泛应用。既往研究表明,肿瘤大小、内部回声特征、基底宽窄及彩色多普勒血流成像(color Doppler flow imaging,CDFI)信号等超声图像特征与膀胱癌的生物学行为及病理分级密切相关[6-7]。然而,单一超声影像学特征受限于操作者经验、图像分辨率及肿瘤本身异质性,术前准确评估肿瘤病理分级的能力仍有限[8]

    越来越多的研究证实,系统性炎症反应可影响肿瘤微环境,进而干预肿瘤发生、进展及转移过程[9]。在膀胱癌领域,已有研究证实中性粒细胞与淋巴细胞比值(neutrophil-to-lymphocyte ratio,NLR)、血小板与淋巴细胞比值(platelet-to-lymphocyte ratio,PLR)、淋巴细胞与单核细胞比值(nymphocyte-to-monocyte ratio,LMR)及系统性免疫炎症指数(systemic immune-inflammation index,SII)等多种血液炎症指标的升高与肿瘤病理分级升高及患者预后不良密切相关[10]。然而,现有研究主要采用单因素分析探讨单一炎症指标与肿瘤分级或预后之间的关联[11],尚缺乏将多种炎症指标和影像学特征整合构建系统性预测模型的深入研究。同时,临床上亦缺乏能够无创评估辅助膀胱癌病理分级的、可直观应用的可视化工具(如列线图)[12-13]。因此,基于超声特征及血液炎症指标整合构建术前预测模型具有重要的研究价值和潜在的临床应用前景。

    本研究回顾性分析了我院既往UCB患者的临床资料、超声特征及血液炎症指标,采用单因素及多因素logistic回归分析筛选独立预测因素,并基于此构建术前无创预测UCB病理分级的列线图模型。通过ROC曲线、校准曲线及决策曲线分析(decision curve analysis,DCA)对模型的判别能力、校准性能及临床实用性进行评估,期望为UCB患者术前病理分级的早期识别及个体化治疗决策提供参考依据。

    本研究为回顾性分析,通过海军军医大学第二附属医院医学伦理委员会审核批准。纳入2017年1月至2024年12月在海军军医大学第二附属医院行膀胱超声检查并经病理确诊为UCB的患者。

    纳入标准:(1)术前超声检查图像清晰,能够完整显示膀胱壁结构及病变特征;(2)超声检查提示膀胱腔内存在占位性病变,符合尿路上皮来源影像学表现;(3)经手术切除或膀胱镜检查联合活检病理确诊为UCB,并明确肿瘤病理分级(低级别或高级别);(4)术前3 d内完成血常规检查,包括中性粒细胞、淋巴细胞、血小板计数,数据完整可用于炎症指标NLR、PLR等的计算;(5)术前未接受放射治疗、化学治疗、免疫治疗等抗肿瘤干预;(6)临床资料齐全,包括超声影像、血液学指标及完整病理资料。

    排除标准:(1)超声图像质量差,无法清晰评估膀胱壁及病变特征者;(2)病理结果为膀胱非尿路上皮来源肿瘤(如鳞状细胞癌、腺癌)或良性病变(如乳头状瘤)者;(3)病变来源于其他脏器肿瘤膀胱侵犯(如前列腺癌、宫颈癌膀胱转移)者;(4)既往接受盆腔放射治疗或膀胱部分切除术者;(5)合并活动性感染(如肺炎、尿路感染)或血液系统疾病(如白血病、骨髓异常增生综合征)者;(6)术前接受新辅助治疗(化射治疗、免疫治疗等)者;(7)临床资料缺失或关键信息(超声影像、血液指标、病理结果)不全者。

    采用Mindray DC-70 Pro超声诊断仪(C5-2E探头,频率2~5 MHz)及Mindray Resona 7超声诊断仪(SC6-1U探头,频率1~6 MHz)进行超声检查,仪器参数设置恒定:聚焦深度约为5.0 cm,热指数为0.0,机械指数为1.3。CDFI统一设置如下参数:脉冲重复频率为700~1 000 Hz,彩色增益调节至不出现背景噪声,滤波设置为中等水平,并采用低速血流模式。血流信号阳性定义为肿瘤内部或边缘可见稳定彩色血流灶,排除运动伪影与生理性干扰。检查前嘱患者适量饮水,待膀胱充盈后取仰卧位行经腹部超声检查,多切面扫查膀胱壁。记录病灶的部位、数目、最大径线、内部回声、边缘形态、基底特征、钙化情况及CDFI信号。每例患者的超声图像由2名具有10年以上工作经验的超声科主治医师采用双盲法独立分析。

    收集患者临床病史,同时通过医院电子病历系统收集患者治疗前的血常规检查数据,并计算相关炎症指标,包括SII(血小板计数×中性粒细胞计数/淋巴细胞计数)、NLR、PLR和LMR。本研究纳入患者的实验室检查(包括炎症指标)均在入院当天完成(入院24 h内),以确保炎症指标的测定结果与超声检查和病理诊断的时间一致性。

    采用SPSS 26.0软件及R 4.5.0软件进行数据分析。正态分布的计量资料以x±s表示,采用独立样本t检验进行组间比较;非正态分布的计量资料以MQ1Q3)表示,采用Mann-Whitney U检验进行组间比较;计数资料以例数和百分数表示,采用χ2检验或Fisher确切概率法进行组间比较。

    在变量筛选阶段,首先对临床特征、超声特征及炎症指标进行单因素logistic回归分析,将其中P<0.05的变量纳入多因素logistic回归分析,以确定的预测因子建立预测模型并绘制列线图。

    通过ROC曲线及AUC值评估列线图模型的区分能力;利用自助法(Bootstrap)重复抽样1 000次绘制校准曲线,检验模型预测概率与实际观察结果的一致性。考虑样本类别不均衡(高级别样本约占85%,低级别样本约占15%)的影响,采用分层Bootstrap方法进行内部验证:按病理分级分层抽样后重复构建对应模型并计算AUC值,循环1 000次,报告平均AUC值及其95%CI作为稳健性评估依据。采用DCA评估模型在不同风险阈值下的净获益。多组间ROC曲线AUC值的分析采用DeLong检验。为控制多重比较错误率,采用Bonferroni法校正(阈值0.017)。所有统计检验均为双侧检验,检验水准(α)为0.05。

    最终纳入471例患者,其中高级别UCB组401例、低级别UCB组70例。高级别组男339例(84.5%)、女62例(15.5%),平均年龄为(67.08±11.91)岁;低级别组男53例(75.7%)、女17例(24.3%),平均年龄为(52.81±16.12)岁。高级别组病灶最大径为(1.70±0.85)cm,低级别组为(1.57±0.52)cm。所有病例均经手术切除或膀胱镜检查联合活检病理确诊,未见其他组织类型或混合型肿瘤成分。本研究未纳入腺癌、鳞状细胞癌、肉瘤样癌及其他非尿路上皮来源的膀胱肿瘤病例。

    单因素logistic回归分析结果显示,在一般临床资料方面,高级别组男性UCB患者占比(OR=1.75,95%CI 1.42~3.23)、肉眼血尿发生率(OR=3.01,95%CI 1.77~5.13)、平均年龄(OR=1.09,95%CI 1.06~1.11)均高于低级别组(均P<0.05)。在超声特征方面,高级别组病灶位于三角区(OR=4.17,95%CI 2.46~7.07),血流信号检出率较高(OR=3.23,95%CI 1.92~5.44),单发病灶比例较高(OR=2.40,95%CI 1.10~5.24),宽基底病灶比例较高(OR=4.39,95%CI 1.22~15.81)。在炎症指标方面,NLR在高级别组UCB患者中高于低级别组(OR=1.07,95%CI 1.02~1.25),其余指标如中性粒细胞计数、淋巴细胞计数、单核细胞计数、血小板计数、PLR、LMR、SII在两组间的差异均无统计学意义(均P>0.05)。见表 1

    表  1  高、低级别UCB患者一般临床资料、超声特征及炎症指标的单因素logistic回归分析
    Table  1  Univariate logistic regression analysis of clinical data, ultrasonographic features and inflammatory indicatorsin high- and low-grade UCB patients
    Variable High-grade group N=401 Low-grade group N=70 OR (95%CI) P value
    Basic clinical data
      Gender, n
        Male 339 53 1.75 (1.42, 3.23) 0.041
        Female 62 17
      Macroscopic hematuria, n
        Yes 321 40 3.01 (1.77, 5.13) <0.001
        No 80 30
      Age/year, x±s 67.08±11.91 52.81±16.12 1.09 (1.06, 1.11) <0.001
    Ultrasonographic feature
      Located in trigone, n
        Yes 295 42 4.17 (2.46, 7.07) <0.001
        No 106 28
      Maximum diameter/cm, x±s 1.70±0.85 1.57±0.52 1.12 (0.91, 1.38) 0.085
      Blood flow signal, n
        Yes 279 29 3.23 (1.92, 5.44) <0.001
        No 122 41
      Number of lesions, n
        Single 375 60 2.40 (1.10, 5.24) 0.046
        Multiple 26 10
      Echogenicity, n
        Hypoechoic 390 68 0.49 (0.21, 1.17) 0.108
        Slightly hyperechoic 11 2
      Margin characteristic, n
        Regular 104 26 0.59 (0.29, 1.21) 0.075
        Irregular 297 44
      Lesion base, n
        Broad 386 37 4.39 (1.22, 15.81) 0.024
        Narrow 15 33
      Calcification, n
        Yes 132 16 0.58 (0.28, 1.21) 0.144
        No 269 54
    Inflammatory indicator
      Neutrophil/(L-1, ×109), M (Q1, Q3) 4.36 (3.31, 5.31) 4.20 (3.62, 5.08) 1.10 (0.85, 1.42) 0.277
      Lymphocyte/(L-1, ×109), M (Q1, Q3) 1.60 (1.33, 1.88) 1.58 (1.50, 1.69) 1.08 (0.85, 1.37) 0.097
      Monocyte/(L-1, ×109), M (Q1, Q3) 0.51 (0.42, 0.59) 0.49 (0.42, 0.56) 1.02 (0.61, 1.71) 0.603
      Platelet/(L-1, ×109), M (Q1, Q3) 219.86 (189.32, 256.52) 228.26 (194.80, 259.85) 1.00 (0.99, 1.01) 0.496
      NLR, M (Q1, Q3) 2.99 (1.87, 4.00) 2.66 (1.46, 3.66) 1.07 (1.02, 1.25) 0.045
      PLR, M (Q1, Q3) 109.14 (77.37, 142.11) 98.08 (73.01, 122.11) 1.02 (0.99, 1.05) 0.327
      LMR, M (Q1, Q3) 4.73 (3.83, 5.57) 5.16 (4.13, 6.52) 0.93 (0.80, 1.10) 0.335
      SII/(L-1, ×109), M (Q1, Q3) 323.54 (282.48, 370.24) 315.69 (288.41, 338.24) 1.02 (0.77, 1.03) 0.256
    UCB: Urothelial carcinoma of bladder; NLR: Neutrophil-to-lymphocyte ratio; PLR: Platelet-to-lymphocyte ratio; LMR: Lymphocyte-to-monocyte ratio; SII: Systemic immune-inflammation index; OR: Odds ratio; 95%CI: 95% confidence interval.

    将单因素logistic回归分析中有统计学意义的变量纳入多因素logistic回归分析,结果显示性别(OR=2.68,95%CI 1.21~5.97,P=0.016)、年龄(OR=1.08,95%CI 1.05~1.11,P<0.001)、肉眼血尿(OR=3.19,95%CI 1.55~6.57,P=0.002)、病灶位于三角区(OR=4.59,95%CI 2.29~9.21,P<0.001)、血流信号阳性(OR=2.87,95%CI 1.45~5.68,P=0.002)和NLR(OR=1.03,95%CI 1.01~1.05,P=0.003)均为UCB高级别病理分级的独立预测因子。基于多因素logistic回归分析结果,按照回归系数实际比例进行赋值,构建高级别UCB列线图预测模型(图 1)。

    图  1  基于多因素logistic回归分析结果构建的高级别UCB列线图预测模型
    Fig.  1  Nomogram predictive model for high-grade UCB based on multivariate logistic regression analysis
    The nomogram incorporates 6 variables: gender, age, macroscopic hematuria, lesion location, blood flow signal, and NLR. The total score corresponds to the predicted probability of having high-grade UCB. UCB: Urothelial carcinoma of bladder; NLR: Neutrophil-to-lymphocyte ratio.
    下载: 全尺寸图片
    2.4.1   ROC曲线分析

    分别基于一般临床资料(性别、年龄、肉眼血尿,模型1)、一般临床资料+超声特征(性别、年龄、肉眼血尿、病灶位于三角区、血流信号阳性,模型2)和一般临床资料+超声特征+炎症因子(性别、年龄、肉眼血尿、病灶位于三角区、血流信号阳性、NLR,模型3)构建了3种预测模型,并分别绘制ROC曲线评估其区分能力。结果显示,模型1预测高级别UCB的AUC值为0.799,模型2为0.856,模型3为0.892(图 2)。

    图  2  3种高级别UCB预测模型的ROC曲线
    Fig.  2  ROC curves of 3 predictive models for high-grade UCB
    Model 1 was constructed based on basic clinical data, model 2 was constructed based on basic clinical data+ultrasonographic features, and model 3 was constructed based on basic clinical data+ultrasonographic features+inflammatory indicators. UCB: Urothelial carcinoma of bladder; ROC: Receiver operating characteristic; AUC: Area under curve.
    下载: 全尺寸图片
    2.4.2   校准曲线分析

    采用自助法重复抽样1 000次绘制各预测模型的校准曲线,模型1(图 3A)的校准曲线与理想线存在一定偏离,模型2(图 3B)的校准曲线拟合度较模型1提高,模型3(图 3C)的校准曲线与理想线拟合较好,3种模型均在一定范围内表现出不同程度的一致性。

    图  3  3种高级别UCB预测模型的校准曲线
    Fig.  3  Calibration curves of 3 predictive models for high-grade UCB
    A: Model 1 (based on basic clinical data); B: Model 2 (based on basic clinical data+ultrasonographic features); C: Model 3 (based on basic clinical data+ultrasonographic features+inflammatory indicators). The red dashed line represents the ideal calibration, the blue solid line represents the predicted probabilities from the models and the black dashed line indicates bias-corrected calibration. UCB: Urothelial carcinoma of bladder.
    下载: 全尺寸图片
    2.4.3   DCA结果

    通过DCA评估各预测模型在不同风险阈值下的临床应用价值,结果如图 4所示。模型1在高风险阈值>0.5时提供净获益,模型2整体净获益高于模型1,而模型3在较广泛阈值区间内均表现出最高的标准化净获益。随着代价-获益比从1∶100到100∶1变化,模型3相较于“全治疗”或“全不治疗”策略在临床决策中能获得更大的净获益。

    图  4  3种高级别UCB预测模型的决策曲线分析
    Fig.  4  Decision curve analysis of 3 predictive models for high-grade UCB
    Model 1 was constructed based on basic clinical data, model 2 was constructed based on basic clinical data+ultrasonographic features, and model 3 was constructed based on basic clinical data+ultrasonographic features+inflammatory indicators. UCB: Urothelial carcinoma of bladder.
    下载: 全尺寸图片
    2.4.4   分层Bootstrap验证结果及各预测模型的性能比较结果

    与模型1相比,模型2预测高级别UCB的AUC值提高0.058(95%CI 0.025~0.097),模型3提高0.110(95%CI 0.065~0.159);模型3较模型2亦提高0.052(95%CI 0.018~0.092)。上述95%CI均不包含0,提示炎症指标的加入进一步提高了模型的判别力(表 2)。在分层Bootstrap验证基础上进一步采用DeLong检验对3个预测模型的AUC值进行两两比较,结果显示,各预测模型间AUC值差异均有统计学意义(均P<0.017,经Bonferroni法校正)。

    表  2  各预测模型的分层Bootstrap验证结果(1 000次迭代)
    Table  2  Results of stratified Bootstrap validation (1 000 iterations) for each predictive model
    Model Mean AUC (95%CI) ΔAUC vs model 1 (95%CI) ΔAUC vs model 2 (95%CI)
    Model 1 0.802 (0.743, 0.856)
    Model 2 0.860 (0.805, 0.911) 0.058 (0.025, 0.097)
    Model 3 0.912 (0.870, 0.950) 0.110 (0.065, 0.159) 0.052 (0.018, 0.092)
    Model 1 was constructed based on basic clinical data, model 2 was constructed based on basic clinical data+ultrasonographic features, and model 3 was constructed based on basic clinical data+ultrasonographic features+inflammatory indicators. AUC: Area under curve; 95%CI: 95% confidence interval.

    选取2例UCB患者进行列线图模型临床应用验证。第1例患者女性,50岁,出现肉眼血尿,超声检查显示膀胱非三角区占位,CDFI未见明显血流信号(图 5A5B),NLR为2.0,模型总评分为54分,对应的高级别UCB概率为30%,术后病理结果为低级别乳头状尿路上皮癌。第2例患者男性,55岁,出现肉眼血尿,超声检查显示膀胱非三角区占位,CDFI内见点片状血流信号(图 5C5D),NLR为2.6,模型总评分为79分,对应的高级别UCB概率>90%,术后病理结果为高级别浸润性尿路上皮癌。

    图  5  2例UCB患者的超声图像特征
    Fig.  5  Ultrasonographic features of 2 patients with UCB
    A, B: Ultrasonography of patient 1 showed a lesion in the non-trigone area of the bladder with no obvious blood flow signal on CDFI; C, D: Ultrasonography of patient 2 showed a lesion in the non-trigone area of the bladder with patchy blood flow signals visible on CDFI. UCB: Urothelial carcinoma of bladder; CDFI: Color Doppler flow imaging.
    下载: 全尺寸图片

    UCB是泌尿系统最常见的恶性肿瘤之一,具有高复发率和进展倾向,严重影响患者的生活质量与长期预后[3, 14]。高级别UCB通常伴有较高的转移风险,往往需行根治性膀胱切除术,低级别病变则多通过经尿道肿瘤切除术进行干预[15-16]。因此,明确的病理分级对于制定合理的治疗方案和评估患者预后至关重要。尽管膀胱镜检查联合组织活检仍为当前UCB病理分级的标准方法,但由于其侵袭性强、成本高及患者依从性差,限制了在术前的广泛应用[5]。在影像学特征及炎症指标与肿瘤生物学行为密切相关的背景下,本研究尝试整合UCB患者的一般临床资料、超声影像特征及血液炎症指标,构建术前无创UCB病理分级预测模型,并以列线图形式实现可视化,旨在为临床个体化治疗决策提供参考依据。

    在本研究构建的术前UCB病理分级预测模型中,性别、年龄、肉眼血尿、病灶位于三角区、血流信号阳性和NLR被确认为独立预测因素。这些因素从生物学和临床机制上体现了肿瘤行为和宿主反应的密切关联[17]。在既往一些针对UCB患者性别、年龄等基本临床特征的研究中,Kaynar等[18]指出男性膀胱癌患者更易发生肌层浸润性病变,Viers等[19]则发现高龄患者接受根治性膀胱切除术时更易出现高级别及高分期病理结果,且癌症特异性生存率显著降低。这些研究提示性别与年龄在膀胱癌进展过程中具有重要作用,本研究结果提示男性患者发生高级别UCB的风险高于女性(OR=2.68,95%CI 1.21~5.97,P=0.016),年龄增加亦被确认为独立危险因素(OR=1.08,95%CI 1.05~1.11,P<0.001)。性别与年龄作为基础临床变量的重要性提示宿主因素在肿瘤生物学进程中的关键作用。肉眼血尿作为膀胱癌最常见的首发表现,在本研究中与高级别UCB显著关联(OR=3.19,95%CI 1.55~6.57,P=0.002)。依据欧洲泌尿外科协会指南及临床共识,肉眼血尿通常提示更严重的膀胱病变,尤其在中老年患者中应高度警惕潜在的高级别肿瘤[20]

    在超声影像学特征方面,本研究结果显示超声提示病灶位于膀胱三角区的患者病理分级为高级别的风险高于非三角区病灶者(OR=4.59,95%CI 2.29~9.21,P<0.001)。这一现象可能与膀胱三角区血供丰富、局部屏障功能薄弱、易于肿瘤浸润的解剖学特性有关[21]。此外,本研究还观察到,肿瘤内部血流信号阳性亦显著关联于高级别UCB(OR=2.87,95%CI 1.45~5.68,P=0.002),这与既往关于血流信号与膀胱癌侵袭性相关的研究[22-23]结论一致。

    在系统性炎症指标方面,NLR作为一种简单易得的血液生物标志物,在本研究中展现出稳定的预测能力,NLR每增加1.0高级别UCB风险增加约3%(OR=1.03,95%CI 1.01~1.05,P=0.003)。这一发现与Ivanić等[24]提出的术前NLR升高与非肌层浸润性膀胱癌患者的复发风险相关,以及Potretzke等[25]研究中高NLR预测术后病理分级提高的结果一致。可见,本研究筛选出的独立预测因子在机制层面具有合理性,在现有文献中亦得到充分验证,进一步支持基于一般临床资料、超声特征及血液炎症指标建立的术前无创UCB病理分级预测模型具有科学性与临床应用潜力。

    本研究发现,UCB患者超声显示的病灶数目(OR=2.40,95%CI 1.10~5.24)和基底宽窄(OR=4.39,95%CI 1.22~15.81)在单因素logistic回归分析中也有统计学意义(均P<0.05),但在多因素logistic回归分析中未能作为独立预测因子保留。可能是由于病灶数目与病灶体积、位置等存在一定程度的共线性,基底宽窄的超声评估也存在较高主观性,这均可能影响变量的稳定性[26-27]。这一现象提示在多因素建模过程中,应综合考量统计显著性、变量独立性及可操作性,优化变量筛选策略[28]

    本研究基于一般临床资料、超声特征及血液炎症指标构建了用于术前无创预测UCB病理分级的多因素列线图模型。结果显示,随着超声特征和炎症指标的逐步纳入,模型的预测性能持续提升。单纯基于一般临床资料的模型(模型1)AUC值为0.799,联合超声特征后的模型(模型2)AUC值升高至0.856,综合纳入一般临床资料、超声特征及血液炎症指标的最终模型(模型3)AUC值进一步提高至0.892。校准曲线提示模型预测概率与实际观察结果具有良好一致性,DCA也表明在临床常见风险阈值范围(10%~90%)内模型能够获得最大的标准化净获益。本研究所构建的预测工具区别于传统单一依赖影像学特征或血液指标的方法,而是基于无创、低成本且常规易得的检查项目,具有良好的普适性与患者接受度,列线图形式的可视化也进一步提升了术前风险评估的便捷性与直观性,具有较高的应用价值与转化潜力,有助于辅助临床精准制定个体化治疗方案。

    本研究通过分层Bootstrap验证保证每次重采样均包含足量低级别病例,有效降低了类别失衡导致的潜在过拟合风险。模型3在原始样本中的AUC值为0.892,分层Bootstrap验证(1 000次)结果中平均AUC值为0.912(95%CI 0.870~0.950),这种差异归因于抽样波动,也证明模型判别能力稳健,低级别病例并未显著拉低模型整体判别力,且炎症相关指标(NLR)的引入在统计学上可提升模型性能。这进一步印证了炎症反应在UCB病理分级预测中的独立价值。此外,对各预测模型AUC值的DeLong检验结果也支持将超声特征及炎症指标联合纳入。

    尽管本研究建立的模型展现出较高的预测准确性和临床应用潜力,但仍存在一定局限性。首先,本研究为单中心、回顾性设计,样本量有限,尚缺乏外部验证,需在多中心、大样本人群中进一步检验模型的稳定性与普适性。其次,本研究中高级别与低级别UCB患者的样本量存在明显不均(401例vs 70例),这可能导致模型对低级别UCB的预测性能被高估,从而在实际应用中出现一定程度的类别不平衡偏倚,未来应进一步优化分层抽样设计或引入模型校正策略以提升泛化能力。此外,超声评估的操作者依赖性及设备差异亦可能对模型推广应用有一定影响。未来,结合CT、MRI功能成像特征及分子生物标志物,融合人工智能技术,构建多模态、智能化的预测系统,并通过持续优化与外部验证,有望进一步提升术前无创UCB病理分级评估的准确性与临床应用价值,助力精准医疗在UCB诊疗中的推广与实践。

    综上所述,本研究表明,整合一般临床资料(性别、年龄、肉眼血尿)、超声特征(病灶位于三角区、血流信号阳性)及血液炎症指标(NLR)可有效构建UCB术前无创病理分级预测模型,该模型预测性能良好,有望作为辅助判断UCB病理分级、优化术前治疗决策的重要工具。本研究为UCB术前无创、精准分级提供了数据支撑与实践依据,未来可结合人工智能图像识别与智能建模技术提升模型在多中心临床中的适用性与自动化水平。

  • 图  1   基于多因素logistic回归分析结果构建的高级别UCB列线图预测模型

    Fig.  1   Nomogram predictive model for high-grade UCB based on multivariate logistic regression analysis

    The nomogram incorporates 6 variables: gender, age, macroscopic hematuria, lesion location, blood flow signal, and NLR. The total score corresponds to the predicted probability of having high-grade UCB. UCB: Urothelial carcinoma of bladder; NLR: Neutrophil-to-lymphocyte ratio.

    下载: 全尺寸图片

    图  2   3种高级别UCB预测模型的ROC曲线

    Fig.  2   ROC curves of 3 predictive models for high-grade UCB

    Model 1 was constructed based on basic clinical data, model 2 was constructed based on basic clinical data+ultrasonographic features, and model 3 was constructed based on basic clinical data+ultrasonographic features+inflammatory indicators. UCB: Urothelial carcinoma of bladder; ROC: Receiver operating characteristic; AUC: Area under curve.

    下载: 全尺寸图片

    图  3   3种高级别UCB预测模型的校准曲线

    Fig.  3   Calibration curves of 3 predictive models for high-grade UCB

    A: Model 1 (based on basic clinical data); B: Model 2 (based on basic clinical data+ultrasonographic features); C: Model 3 (based on basic clinical data+ultrasonographic features+inflammatory indicators). The red dashed line represents the ideal calibration, the blue solid line represents the predicted probabilities from the models and the black dashed line indicates bias-corrected calibration. UCB: Urothelial carcinoma of bladder.

    下载: 全尺寸图片

    图  4   3种高级别UCB预测模型的决策曲线分析

    Fig.  4   Decision curve analysis of 3 predictive models for high-grade UCB

    Model 1 was constructed based on basic clinical data, model 2 was constructed based on basic clinical data+ultrasonographic features, and model 3 was constructed based on basic clinical data+ultrasonographic features+inflammatory indicators. UCB: Urothelial carcinoma of bladder.

    下载: 全尺寸图片

    图  5   2例UCB患者的超声图像特征

    Fig.  5   Ultrasonographic features of 2 patients with UCB

    A, B: Ultrasonography of patient 1 showed a lesion in the non-trigone area of the bladder with no obvious blood flow signal on CDFI; C, D: Ultrasonography of patient 2 showed a lesion in the non-trigone area of the bladder with patchy blood flow signals visible on CDFI. UCB: Urothelial carcinoma of bladder; CDFI: Color Doppler flow imaging.

    下载: 全尺寸图片

    表  1   高、低级别UCB患者一般临床资料、超声特征及炎症指标的单因素logistic回归分析

    Table  1   Univariate logistic regression analysis of clinical data, ultrasonographic features and inflammatory indicatorsin high- and low-grade UCB patients

    Variable High-grade group N=401 Low-grade group N=70 OR (95%CI) P value
    Basic clinical data
      Gender, n
        Male 339 53 1.75 (1.42, 3.23) 0.041
        Female 62 17
      Macroscopic hematuria, n
        Yes 321 40 3.01 (1.77, 5.13) <0.001
        No 80 30
      Age/year, x±s 67.08±11.91 52.81±16.12 1.09 (1.06, 1.11) <0.001
    Ultrasonographic feature
      Located in trigone, n
        Yes 295 42 4.17 (2.46, 7.07) <0.001
        No 106 28
      Maximum diameter/cm, x±s 1.70±0.85 1.57±0.52 1.12 (0.91, 1.38) 0.085
      Blood flow signal, n
        Yes 279 29 3.23 (1.92, 5.44) <0.001
        No 122 41
      Number of lesions, n
        Single 375 60 2.40 (1.10, 5.24) 0.046
        Multiple 26 10
      Echogenicity, n
        Hypoechoic 390 68 0.49 (0.21, 1.17) 0.108
        Slightly hyperechoic 11 2
      Margin characteristic, n
        Regular 104 26 0.59 (0.29, 1.21) 0.075
        Irregular 297 44
      Lesion base, n
        Broad 386 37 4.39 (1.22, 15.81) 0.024
        Narrow 15 33
      Calcification, n
        Yes 132 16 0.58 (0.28, 1.21) 0.144
        No 269 54
    Inflammatory indicator
      Neutrophil/(L-1, ×109), M (Q1, Q3) 4.36 (3.31, 5.31) 4.20 (3.62, 5.08) 1.10 (0.85, 1.42) 0.277
      Lymphocyte/(L-1, ×109), M (Q1, Q3) 1.60 (1.33, 1.88) 1.58 (1.50, 1.69) 1.08 (0.85, 1.37) 0.097
      Monocyte/(L-1, ×109), M (Q1, Q3) 0.51 (0.42, 0.59) 0.49 (0.42, 0.56) 1.02 (0.61, 1.71) 0.603
      Platelet/(L-1, ×109), M (Q1, Q3) 219.86 (189.32, 256.52) 228.26 (194.80, 259.85) 1.00 (0.99, 1.01) 0.496
      NLR, M (Q1, Q3) 2.99 (1.87, 4.00) 2.66 (1.46, 3.66) 1.07 (1.02, 1.25) 0.045
      PLR, M (Q1, Q3) 109.14 (77.37, 142.11) 98.08 (73.01, 122.11) 1.02 (0.99, 1.05) 0.327
      LMR, M (Q1, Q3) 4.73 (3.83, 5.57) 5.16 (4.13, 6.52) 0.93 (0.80, 1.10) 0.335
      SII/(L-1, ×109), M (Q1, Q3) 323.54 (282.48, 370.24) 315.69 (288.41, 338.24) 1.02 (0.77, 1.03) 0.256
    UCB: Urothelial carcinoma of bladder; NLR: Neutrophil-to-lymphocyte ratio; PLR: Platelet-to-lymphocyte ratio; LMR: Lymphocyte-to-monocyte ratio; SII: Systemic immune-inflammation index; OR: Odds ratio; 95%CI: 95% confidence interval.

    表  2   各预测模型的分层Bootstrap验证结果(1 000次迭代)

    Table  2   Results of stratified Bootstrap validation (1 000 iterations) for each predictive model

    Model Mean AUC (95%CI) ΔAUC vs model 1 (95%CI) ΔAUC vs model 2 (95%CI)
    Model 1 0.802 (0.743, 0.856)
    Model 2 0.860 (0.805, 0.911) 0.058 (0.025, 0.097)
    Model 3 0.912 (0.870, 0.950) 0.110 (0.065, 0.159) 0.052 (0.018, 0.092)
    Model 1 was constructed based on basic clinical data, model 2 was constructed based on basic clinical data+ultrasonographic features, and model 3 was constructed based on basic clinical data+ultrasonographic features+inflammatory indicators. AUC: Area under curve; 95%CI: 95% confidence interval.
  • [1] SOLOMON J P, HANSEL D E. Prognostic factors in urothelial carcinoma of the bladder: histologic and molecular correlates[J]. Adv Anat Pathol, 2015, 22(2): 102-112. DOI: 10.1097/PAP.0000000000000050.
    [2] LEE C H, TAN C H, DE CASTRO FARIA S, et al. Role of imaging in the local staging of urothelial carcinoma of the bladder[J]. AJR Am J Roentgenol, 2017, 208(6): 1193-1205. DOI: 10.2214/AJR.16.17114.
    [3] MAYR R, FRITSCHE H M, PYCHA A, et al. Radical cystectomy and the implications of comorbidity[J]. Expert Rev Anticancer Ther, 2014, 14(3): 289-295. DOI: 10.1586/14737140.2014.868775.
    [4] WU S, LI R, JIANG Y, et al. Liquid biopsy in urothelial carcinoma: detection techniques and clinical applications[J]. Biomed Pharmacother, 2023, 165: 115027. DOI: 10.1016/j.biopha.2023.115027.
    [5] BOLENZ C, WEST A M, ORTIZ N, et al. Urinary cytology for the detection of urothelial carcinoma of the bladder: a flawed adjunct to cystoscopy[J]. Urol Oncol, 2013, 31(3): 366-371. DOI: 10.1016/j.urolonc.2011.01.017.
    [6] AKINTELURE D, AKINTELURE S, AL-IBRAHEEM H A. Emerging role of multiparametric MRI in the staging of bladder cancer: insights from the BladderPath trial[J]. Cureus, 2025, 17(7): e88623. DOI: 10.7759/cureus.88623.
    [7] 张静, 梁羽, 范尔兮, 等. 基于超声特征的列线图模型鉴别诊断膀胱隆起样病变良恶性的价值[J]. 中国医学影像学杂志, 2024, 32(8): 841-844. DOI: 10.3969/j.issn.1005-5185.2024.08.016.
    [8] ALQAHTANI A, BHATTACHARJEE S, ALMOPTI A, et al. Radiomics-based machine learning approach for the prediction of grade and stage in upper urinary tract urothelial carcinoma: a step towards virtual biopsy[J]. Int J Surg, 2024, 110(6): 3258-3268. DOI: 10.1097/JS9.0000000000001483.
    [9] FONTES M S, DE ALMEIDA D V P, CÁRCANO F, et al. Precision medicine for urothelial carcinoma: an international perspective[J]. Urol Oncol, 2024, 42(12): 402-410. DOI: 10.1016/j.urolonc.2023.11.008.
    [10] ZHANG J H, STARR S L, CHAMIE K. Contemporary systemic therapies in urothelial carcinoma[J]. Urology, 2023, 174: 150-158. DOI: 10.1016/j.urology.2023.01.007.
    [11] FUJIWARA Y, KAROL A B, JOSHI H, et al. C-reactive protein (CRP) as a prognostic biomarker in patients with urothelial carcinoma: a systematic review and meta-analysis[J]. Crit Rev Oncol Hematol, 2024, 197: 104352. DOI: 10.1016/j.critrevonc.2024.104352.
    [12] ARITA Y, YOSHIDA S, SHIGETA K, et al. Diagnostic value of the vesical imaging-reporting and data system in bladder urothelial carcinoma with variant histology[J]. Eur Urol Oncol, 2023, 6(1): 99-102. DOI: 10.1016/j.euo.2022.07.006.
    [13] 任乾,曾蜀雄,汤仕杰,等.上尿路尿路上皮癌术后膀胱复发危险因素分析[J].第二军医大学学报,2018,39(5):469-473. DOI: 10.16781/j.0258-879x.2018.05.0469.

    REN Q, ZENG S X, TANG S J, et al. Risk factors of intravesical recurrence after upper-tract urothelial carcinoma surgery[J]. Acad J Sec Mil Med Univ, 2018, 39(5): 469-473. DOI: 10.16781/j.0258-879x.2018.05.0469.
    [14] MCDONALD S, KEANE K G, GAUCI R, et al. Nuclear medicine and molecular imaging in urothelial cancer: current status and future directions[J]. Cancers, 2025, 17(2): 232. DOI: 10.3390/cancers17020232.
    [15] SERONT E, MACHIELS J P. Molecular biology and targeted therapies for urothelial carcinoma[J]. Cancer Treat Rev, 2015, 41(4): 341-353. DOI: 10.1016/j.ctrv.2015.03.004.
    [16] IKEDA S, HANSEL D E, KURZROCK R. Beyond conventional chemotherapy: emerging molecular targeted and immunotherapy strategies in urothelial carcinoma[J]. Cancer Treat Rev, 2015, 41(8): 699-706. DOI: 10.1016/j.ctrv.2015.06.004.
    [17] CAN C, BASESKIOGLU B, YıLMAZ M, et al. Pretreatment parameters obtained from peripheral blood sample predicts invasiveness of bladder carcinoma[J]. Urol Int, 2012, 89(4): 468-472. DOI: 10.1159/000343278.
    [18] KAYNAR M, YıLDıRıM M E, BADEM H, et al. Bladder cancer invasion predictability based on preoperative neutrophil-lymphocyte ratio[J]. Tumor Biol, 2014, 35(7): 6601-6605. DOI: 10.1007/s13277-014-1889-x.
    [19] VIERS B R, BOORJIAN S A, FRANK I, et al. Pretreatment neutrophil-to-lymphocyte ratio is associated with advanced pathologic tumor stage and increased cancer-specific mortality among patients with urothelial carcinoma of the bladder undergoing radical cystectomy[J]. Eur Urol, 2014, 66(6): 1157-1164. DOI: 10.1016/j.eururo.2014.02.042.
    [20] CATHOMAS R, LORCH A, BRUINS H M, et al. The 2021 updated European association of urology guidelines on metastatic urothelial carcinoma[J]. Eur Urol, 2022, 81(1): 95-103. DOI: 10.1016/j.eururo.2021.09.026.
    [21] GUO C C, LEE S, LEE J G, et al. Molecular profile of bladder cancer progression to clinically aggressive subtypes[J]. Nat Rev Urol, 2024, 21(7): 391-405. DOI: 10.1038/s41585-023-00847-7.
    [22] XUE L Y, LU Q, HUANG B J, et al. Evaluation of renal urothelial carcinoma by contrast-enhanced ultrasonography[J]. Eur J Radiol, 2013, 82(4): e151-7. DOI: 10.1016/j.ejrad.2012.11.001.
    [23] AL-AHMADIE H, NETTO G J. Molecular pathology of urothelial carcinoma[J]. Clin Lab Med, 2024, 44(2): 181-198. DOI: 10.1016/j.cll.2023.08.010.
    [24] IVANIĆD, ÐUG H, JAGODIĆ S, et al. Preoperative neutrophil-to-lymphocyte ratio (NLR) value as a predictor of recurrence of non-muscle-invasive bladder cancer[J]. Int Urol Nephrol, 2025, 57(4): 1143-1149. DOI: 10.1007/s11255-024-04297-x.
    [25] POTRETZKE A, HILLMAN L, WONG K, et al. NLR is predictive of upstaging at the time of radical cystectomy for patients with urothelial carcinoma of the bladder[J]. Urol Oncol, 2014, 32(5): 631-636. DOI: 10.1016/j.urolonc.2013.12.009.
    [26] BALACHANDRAN V P, GONEN M, SMITH J J, et al. Nomograms in oncology: more than meets the eye[J]. Lancet Oncol, 2015, 16(4): e173-80. DOI: 10.1016/S1470-2045(14)71116-7.
    [27] IASONOS A, SCHRAG D, RAJ G V, et al. How to build and interpret a nomogram for cancer prognosis[J]. J Clin Oncol, 2008, 26(8): 1364-1370. DOI: 10.1200/jco.2007.12.9791.
    [28] 朱琚, 聂芳, 王楠, 等. 超声特征联合炎症指标在膀胱尿路上皮癌分级中的应用价值[J]. 肿瘤影像学, 2023, 32(2): 171-177. DOI: 10.19732/j.cnki.2096-6210.2023.02.010.
WeChat 点击查看大图
图(5)  /  表(2)
出版历程
  • 收稿日期:  2025-05-07
  • 接受日期:  2025-08-25

目录

    /

    返回文章
    返回