基于SEER数据库的多原发肺癌临床特征分析及预后预测模型构建

Clinical characteristics analysis and prognostic prediction model construction in multiple primary lung cancer based on the SEER database

目的探讨多原发肺癌(MPLC)的临床特征及预后影响因素,构建预后预测模型。方法回顾性分析基于监测、流行病学和最终结果(SEER)数据库2010年1月至2020年12月收录的经病理检查确诊的MPLC患者的临床资料和预后信息。采用R软件按照7∶3比例将患者随机分为训练集和验证集。采用Kaplan-Meier法绘制生存曲线,组间比较采用log-rank检验。采用单因素及多因素Cox比例风险模型筛选训练集中MPLC患者总生存(OS)的独立影响因素,并依此绘制预测患者3、5、8年生存率的列线图。在训练集和验证集中,以实际生存情况为金标准,绘制依据构建的模型判断患者3、5、8年OS率的受试者工作特征(ROC)曲线,获得曲线下面积(AUC),采用R软件分析模型C指数,绘制模型预测3、5、8年OS率与实际OS率的校准曲线。结果共纳入5495例MPLC患者,训练集3846例,验证集1649例。训练集和验证集不同年龄及AJCC分期患者构成差异均有统计学意义(均P<0.05),其他临床病理特征比较差异均无统计学意义(P>0.05)。多因素Cox回归分析结果显示,男性(与女性相比,HR=1.256,95%CI:1.144~1.379,P<0.001),年龄≥70岁(与50~59岁相比,HR=1.201,95%CI:1.030~1.400,P=0.019),FPLC病理类型为鳞状细胞癌、其他类型(与腺癌相比,HR=1.275,95%CI:1.137~1.431,P<0.001;HR=1.208,95%CI:1.041~1.403,P=0.013),SPLC病理类型为鳞状细胞癌、小细胞肺癌、其他类型(与腺癌相比,HR=1.270,95%CI:1.121~1.440,P<0.001;HR=1.978,95%CI:1.642~2.384,P<0.001;HR=1.246,95%CI:1.090~1.424,P=0.001),AJCC分期Ⅲ、Ⅳ期(与Ⅰ期相比,HR=1.645,95%CI:1.447~1.869,P<0.001;HR=2.078,95%CI:1.669~2.587,P<0.001),FPLC未进行手术(与进行手术相比,HR=1.263,95%CI:1.038~1.536,P=0.020),SPLC未进行手术(进行手术比未进行手术,HR=0.680,95%CI:0.579~0.799,P<0.001),FPLC未进行淋巴结清扫、仅清扫1~3个区域淋巴结(与清扫≥4个相比,HR=1.225,95%CI:1.016~1.477,P=0.034;HR=1.314,95%CI:1.103~1.566,P=0.002),FPLC肿瘤长径3~5 cm、>5 cm(与<3 cm相比,HR=1.181,95%CI:1.053~1.324,P=0.005;HR=1.232,95%CI:1.069~1.420,P=0.004),SPLC肿瘤长径3~5 cm、>5 cm(与<3 cm相比,HR=1.560,95%CI:1.362~1.786,P<0.001;HR=1.727,95%CI:1.451~2.054,P<0.001),FPLC未进行化疗(化疗比未化疗或未知,HR=0.744,95%CI:0.655~0.845,P<0.001)是患者OS不良的独立危险因素。Kaplan-Meier生存分析结果显示,训练集中不同性别、种族、年龄、两次肿瘤位置、AJCC分期、两次肺癌病理类型、两次肿瘤长径及两次肿瘤是否手术治疗、两次肿瘤是否化疗的患者OS比较,差异均有统计学意义(均P<0.05)。根据多因素Cox回归分析结果筛选的影响MPLC患者OS的独立因素绘制了预测MPLC 3、5、8年OS率的列线图。ROC曲线分析结果显示,训练集列线图的C指数为0.679(95%CI:0.649~0.701),预测3、5、8年OS率的AUC值分别为0.601、0.595和0.586;验证集C指数为0.678(95%CI:0.633~0.720),预测3、5、8年OS率的AUC值分别为0.643、0.631和0.626。校准曲线显示,列线图模型预测训练集和验证集患者3、5、8年OS率与实际结果有较好的一致性,拟合度高。结论建立的预后模型具有良好的预测价值,可有效评估患者预后。

Objective To explore the clinical characteristics and prognostic influencing factors of multiple primary lung cancer(MPLC),and to construct a prognostic prediction model.Methods The clinical data and prognostic information of MPLC patients diagnosed by pathological examination included in the Surveillance,Epidemiology,and End Results(SEER)database from January 2010 to December 2020 were retrospectively analyzed.Patients were randomly divided into training and validation sets according to a 7:3 ratio using R software.Survival curves were plotted by using the Kaplan‐Meier method and log‐rank test was used for comparison between groups.The independent influencing factors of overall survival(OS)of MPLC patients in the training set were screened using univariate and multivariate Cox proportional hazards models,and accordingly,the nomogram predicting the survival rate of patients at 3,5 and 8 years were plotted.In the training and validation sets,using the actual survival as the gold standard,the receiver operating characteristic(ROC)curves of the constructed models for predicting the patients'3-,5-and 8-year OS rates were plotted,the area under the curve(AUC)was obtained,and C-index of the model was analyzed by using R software.The calibration curves of 3-,5-and 8-year OS rates predicted by the models and the actual OS rates were plotted.Results A total of 5495 MPLC patients were included,3846 in the training set and 1649 in the validation set.The differences in the composition of patients of different ages and AJCC stages between the training and validation sets were statistically significant(both P<0.05),and the differences in the comparison of other clinicopathological characteristics were not statistically significant(all P>0.05).The results of multivariate Cox regression analysis showed that males(compared with females,HR=1.256,95%CI:1.144-1.379,P<0.001),age≥70 years old(compared with 50-59 years old,HR=1.201,95%CI:1.030-1.400,P=0.019),FPLC with pathological types of squamous cell carcinoma or other types(compared with adenocarcinoma,HR=1.275,95%CI:1.137-1.431,P<0.001;HR=1.208,95%CI:1.041-1.403,P=0.013),and SPLC with pathological types of squamous cell carcinoma,small cell lung carcinoma,or other types(compared with adenocarcinoma,HR=1.270,95%CI:1.121-1.440,P<0.001;HR=1.978,95%CI:1.642-2.384,P<0.001;HR=1.246,95%CI:1.090-1.424,P=0.001),and AJCC stage Ⅲ and Ⅳ(compared with stage Ⅰ,HR=1.645,95%CI:1.447-1.869,P<0.001;HR=2.078,95%CI:1.669-2.587,P<0.001),FPLC without operation(compared with operation,HR=1.263,95%CI:1.038-1.536,P=0.020),SPLC without operation(operation vs.no operation,HR=0.680,95%CI:0.579-0.799,P<0.001),FPLC without lymph node dissection or with clearance of 1-3 regional lymph nodes(compared with clearance of≥4,HR=1.225,95%CI:1.016-1.477,P=0.034;HR=1.314,95%CI:1.103-1.566,P=0.002),FPLC with maximum diameter 3-5 cm or>5 cm(compared with<3 cm,HR=1.181,95%CI:1.053-1.324,P=0.005;HR=1.232,95%CI:1.069-1.420,P=0.004),and SPLC with maximum diameter 3-5 cm or>5 cm(compared with<3 cm,HR=1.560,95%CI:1.362-1.786,P<0.001;HR=1.727,95%CI:1.451-2.054,P<0.001),and FPLC without chemotherapy(chemotherapy vs.no chemotherapy or unknown,HR=0.744,95%CI:0.655-0.845,P<0.001)were the independent risk factors of patients'poor OS(all P<0.05).The results of Kaplan-Meier survival analysis showed that the OS of patients with different gender,race,age,two tumor locations,AJCC staging,pathological type of two lung tumors,maximum diameter of two tumors,and whether two tumors were treated surgically or not,and whether two tumors were treated with chemotherapy or not in the training set were compared,and the differences were all statistically significant(all P<0.05).Based on the independent factors affecting the OS of MPLC patients screened by the results of multivariate Cox regression analysis,nomogram predicting the 3-,5-and 8-year OS rates of MPLC were plotted.The results of ROC curve analysis showed that the C-index of the training set's nomogram was 0.679(95%CI:0.649-0.701),and the AUC values for predicting the 3-,5-and 8-year OS rates were 0.601,0.595 and 0.586,respectively;the C-index of the validation set was 0.678(95%CI:0.633-0.720),and the AUC values for predicting 3-,5-and 8-year OS rates were 0.643,0.631 and 0.626,respectively.The calibration curves showed that the 3-,5-and 8-year OS rates of patients predicted by the nomogram models in both the training and validation sets were in good agreement with the actual results with a high goodness-of-fit.Conclusions The established prognostic model has good predictive value and can effectively assess the prognosis of patients.