T1期结直肠癌淋巴结转移的危险因素分析及其列线图预测模型的应用价值

目的探讨T1期结直肠癌淋巴结转移的危险因素及其列线图预测模型的应用价值。方法采用回顾性病例对照研究方法。收集2008年6月至2019年12月复旦大学附属中山医院收治的914例行根治性切除术T1期结直肠癌病人的临床病理资料;男528例,女386例;中位年龄为63岁,年龄范围为25~87岁。观察指标:(1)T1期结直肠癌病人的临床病理资料。(2)随访情况。(3)淋巴结转移的影响因素分析。(4)列线图预测模型的建立及内部验证。病人术后定期随访,术后2年内每3个月随访1次,随后每6个月随访1次,术后随访5年,了解病人的肿瘤复发和生存情况。正态分布的计量资料以x±s表示,组间比较采用t检验;偏态分布的计量资料以M(范围)表示。计数资料以绝对数或百分比表示,组间比较采用χ2检验。采用Kaplan-Meier法计算生存率和绘制生存曲线。采用Log-rank检验进行生存分析。单因素和多因素分析均采用Logistic回归分析。根据多因素分析结果,应用R语言软件建立基于Logistic回归的淋巴结转移概率预测列线图。采用校准度曲线评价模型预测结局发生概率与实际观测概率一致程度,以一致性指数(C-index)表示。采用Bootstrap方法评价模型性能,得出校准度曲线。采用Hosmer-Lemeshow检验计算模型的拟合优度。结果 (1)T1期结直肠癌病人的临床病理资料:914例病人中,直接手术687例,内镜切除后补救手术227例;术后组织病理学检查证实均为pT1NxM0期结直肠癌;肿瘤长径为(2.3±1.2)cm;肿瘤病理学类型腺癌为865例,黏液性腺癌为49例;肿瘤分化程度为高中分化727例,低未分化187例;黏膜下浸润深度≥1 000 μm 633例,<1 000 μm 281例;神经脉管侵犯110例,未受侵犯804例;术中淋巴结清扫数目为13枚(1~48枚);N分期为N0期804例,N1期98例,N2期12例。无围术期死亡病人。(2)随访情况:914例病人中,886例获得术后随访,随访时间为25个月(1~129个月);随访期间肿瘤复发或转移24例。914例病人5年肿瘤累积复发率为4.8%,中位复发时间为17.0个月,肝脏为最常见的肿瘤复发部位,占比为58.3%(14/24)。914例病人5年无复发生存率为95.2%。804例无淋巴结转移病人5年无复发生存率为96.3%,110例有淋巴结转移病人为86.6%,两者比较,差异有统计学意义(χ2=6.83,P<0.05)。(3)淋巴结转移的影响因素分析:单因素分析结果为术前癌胚抗原(CEA)、术前CA19-9、肿瘤分化程度、黏膜下浸润深度、神经脉管侵犯是影响T1期结直肠癌淋巴结转移的相关因素(优势比=2.56、3.25、2.21、2.68、3.39,95%可信区间为1.41~4.67、1.22~8.66、1.43~3.41、1.56~4.88、2.10~5.48,P<0.05)。多因素分析结果显示:术前CEA≥5 μg/L、术前CA19-9≥37 U/mL、肿瘤分化程度为低未分化、黏膜下浸润深度≥1 000 μm、神经脉管侵犯是影响T1期结直肠癌淋巴结转移的独立危险因素(优势比=2.23、3.47、2.01、2.31、2.91,95%可信区间为1.02~4.15、1.08~10.87、1.03~3.27、1.40~4.47、1.64~5.13,P<0.05)。(4)列线图预测模型的建立及内部验证:根据多因素Logistic分析结果,构建预测T1期结直肠癌淋巴结转移列线图模型。术前CEA≥5 μg/L、术前CA19-9≥37 U/mL、肿瘤分化程度为低未分化、黏膜下浸润深度≥1 000 μm、神经脉管侵犯得分分别为59、100、48、67、92分。根据每项临床病理因素评分,加和得总分后评估淋巴结转移概率。绘制受试者工作特征曲线评价列线图模型的淋巴结转移预测能力,其结果显示:列线图预测模型曲线下面积为0.70(95%可信区间为0.64~0.75,P<0.05)。Bootstrap法验证列线图预测模型的预测效能,C-index值为0.70(95%可信区间为0.65~0.75)。校准度曲线显示该列线图模型的预测概率和实际淋巴结转移概率具备较好的一致性。Hosmer-Lemeshow检验计算模型的拟合效果好(χ2=1.61,P>0.05)。结论术前CEA≥5 μg/L、术前CA19-9≥37 U/mL、肿瘤分化程度为低未分化、黏膜下浸润深度≥1 000 μm、神经脉管侵犯是影响T1期结直肠癌淋巴结转移的独立危险因素;以此构建列线图预测模型,可以预测T1期结直肠癌淋巴结转移概率。

Risk factors for lymph node metastasis in T1 colorectal cancer and application value of its nomogram prediction model

Objective To investigate the risk factors for lymph node metastasis in T1 colorectal cancer and application value of its nomogram prediction model.Methods The retrospective case‐control study was conducted.The clinicopathological data of 914 patients with T1 colorectal cancer who underwent radical resection in the Zhongshan Hospital of Fudan University June 2008 to December 2019 were collected.There were 528 males and 386 females,aged from 25 to 87 years,with a median age of 63 years.Observation indicators:(1)clinicopathological data of patients with T1 colorectal cancer;(2)follow‐up;(3)analysis of influencing factors for lymph nodmetastasis;(4)development and internal validation of a nomogram predition model.Patients were regularlly followed up once three months within postoperative 2 years and once six months thereafter to detect tumor recurrence and survival.The endpoint of follow‐up was at postoperative 5 years.Measurement data with normal distribution were represented as Mean±SD,and comparison between groups was analyzed using the t test.Measurement data with skewed distribution were represented as M(range).Count data were described as absolute numbers or percentages,and comparison between groups was analyzed using the chi‐square test.The Kaplan‐Meier method was used to calculate survival rates and draw survival curves.The Log‐rank test was used for survival analysis.Univariate and multivariate analyses were performed using the Logistic regression analysis.Based on results of multivariate analysis,a Logistic regressional nomogram for prediction of lymph node metastasis probability was constructed using R language software.The calibration curve was used to evaluate the consistency between probability predicd by the nomogram model and actual observation probability,which was reprensented by a consistency index.The Bootstrap method was used for evaluation of the model performance to receive the calibration curve.The Hosmer‐Lemeshow test was used to calculate the goodness of fit in model.Results(1)Clinicopathological data of patients with T1 colorectal cancer:687 of 914 patients underwent direct surgery and 227 underwent remedial operation after endoscopic resection.All the 914 patients were confirmed as pT1NxM0 colorectal cancer by pathological examination.The tumor diameter was(2.3±1.2)cm.The pathological catogaries of 914 patients included 865 cases of adenocarcinoma and 49 case mucinous adenocarcinoma.The tumor differentiation degree of 914 patients included 727 cases of high or middle differentiation and 187 cases of low differentiation or undifferentiation.Of the 914 patients,633 cases had submucosal infiltration depth≥1000μm and 281 cases had submucosal infiltration depth<1000μm.Thre were 110 cases with nerve vessel invasion and 804 without nerve vessel invasion.The number of intraoperative lymph node dissection was 13(range,1-48).There were 804 cases in stage N0 of N staging,98 cases in stage N1 and 12 cases in stage N2.There was no perioperative death.(2)Follow‐up:886 of 914 patients were followed up for 25 months(range,1-129 months).During the follow‐up,24 patients had tumor recurrence or metastasis.The 5‐year cumulative tumor recurrence rate of 914 patients was 4.8%and the median recurrence time was 17.0 months.Liver was the main site of tumor recurrence,accounting for 58.3%(14/24).The 5‐year recurrence‐free survival rate of 914 patients was 95.2%.The 5‐year recurrence‐free survival rate was 96.3%of 804 patients without lymph node metastasis,versus 86.6%of 110 patients with lymph node metastasis,showing a significant difference between the two groups(χ²=6.83,P<0.05).(3)Analysis of influencing factors for lymph node metastasis:results of univariate analysis showed that preoperative carcinoembryonic antigen(CEA),preoperative CA19‐9,tumor differentiation degree,submucosal infiltration depth,nerve vessel invasion were related factors for lymph node metastasis in T1 colorectal cancer(odds ratio=2.56,3.25,2.21,2.68,3.39,95%confidence interval as 1.41-4.67,1.22-8.66,1.43-3.41,1.56-4.88,2.10-5.48,P<0.05).Results of multivariate analysis showed that preoperative CEA≥5μg/L,preoperative CA19‐9≥37 U/mL,poor differentiation or undifferentiation,submucosal infiltration depth≥1000μm and nerve vessel invasion were independent risk factors for lymph node metastasis in T1 colorectal cancer(odds ratio=2.23,3.47,2.01,2.31,2.91,95%confidence interval as 1.02-4.15,1.08-10.87,1.03-3.27,1.40-4.47,1.64-5.13,P<0.05).(4)Development and internal validation of a nomogram predition model:based on results of multivariate Logistic analysis,a nomogram prediction model for lymph node metastasis in T1 colorectal cancer was developed.The nomogram score was 59 for preoperative CEA>5μg/L,100 for preoperative CA19‐9≥37 U/mL,48 for poor differentiation or undifferentiation,67 for submucosal infiltration depth≥1000μm and 92 for nerve vessel invasion,respectively.The total of different scores for different clinicopathological factors corresponded to the probability of lymph node metastasis.The receiver operating characteristic curve was drawed to evaluate the predictive performance of nomogram for lymph node metastasis in T1 colorectal cancer,with the area under curve of 0.70(95%confidence interval as 0.64-0.75,P<0.05).The Bootstrap internal validation of predictive performance in the nomogram predition model showed a consistency index of 0.70(95%confidence interval as 0.65-0.75).The calibration chart showed a good consistency between t probability predicted by the nomogram model and actual probability of lymph node metastasis.The Hosmer‐Lemeshow test showed a good fitting effect in model(χ²=1.61,P>0.05).Conclusions Preoperative CEA≥5μg/L,preoperative CA19‐9≥37 U/mL,poor differentiation or undifferentiation,submucosal infiltration depth≥1000μm and nerve vessel invasion are independent risk factors for lymph node metastasis in T1 colorectal cancer.The constructed nomogram model can help predict the probability of lymph node metastasis in T1 colorectal cancer.