中国人群肺腺癌生存概率列线图的建立与验证:一项基于SEER的大型回顾性队列研究

Establishment and validation of nomograms to survival probability of lung adenocarcinoma in Chinese population:a large retrospective cohort study based on SEER

目的使用“监测、流行病学和最终结果”(SEER)数据库构建列线图,以预测中国肺腺癌患者的癌症特异性生存期(CSS)。方法从SEER数据库中初选2000-2020年间17个登记处的7940名中国肺腺癌患者,根据纳入与排除标准最终纳入3304名患者,随机分配(7∶3)到训练集和验证集。列线图通过单因素及多因素Cox回归筛选变量构建,并通过一致性指数(C-Index)、受试者工作特征曲线(ROC)、校准曲线、决策曲线(DCA)和危险分层Kaplan-Meier曲线进行评估。结果列线图是结构良好且经过充分验证的预后图,由8个变量组成:婚姻状态、原发部位、临床分期、T分型(肿瘤大小)、N分型(淋巴结转移)、外科手术、区域淋巴结清扫和放射治疗。训练集C-Index为0.716(CI:0.702~0.730),验证集C-Index为0.697(CI:0.675~0.719)。在1年、3年和5年的时间点,训练集ROC曲线下面积(AUC)分别为0.766、0.808和0.858,验证集AUC分别为0.733、0.789和0.816。校准曲线证实了列线图预测和观察到的生存概率之间具有一致性,而DCA证实了列线图具有较好的临床适用性。根据训练集的预测值中位数进行风险评分,将所有个体分配到高风险组和低风险组,Kaplan-Meier分析显示两组CSS差异有统计学意义(P<0.001)。结论本研究构建并验证了中国人群肺腺癌的预后列线图。本研究构建的模型在临床环境中的应用可以帮助临床医生评估患者的预后并提供高度个性化的治疗。

Objective To build nomograms with the Surveillance,Epidemiology,and End Results(SEER)database.Then predicting the cancer-specific survival(CSS)of lung adenocarcinoma in Chinese people.Methods A total of 7940 patients eligible for recruitment between 2000 and 2020 were selected from 17 registries of the SEER database.According to the inclusion and exclusion criteria,3304 patients were finally enrolled in and randomized(in a 7∶3 ratio)to development sets and validation sets.Nomograms were constructed from variables which screened by univariate or multivariate Cox regression analyses.Nomograms could be evaluated by consistency indices(C-Index),receiver operating characteristic(ROC)curves,calibration curves,decision curve analysis(DCA)and risk stratification Kaplan-Meier survival curve.Results The nomograms were well-structured and well-validated prognostic maps,it constructed from 8 variables:marital status,primary site,clinical stages,T(size of tumour),N(lymph node metastasis),surgery,scope of regional lymph node surgery and radiation.The C-Index of the development sets was 0.716(CI:0.702-0.730)and 0.697(CI:0.675-0.719)for the validation sets.At these points in time of 1 year,3 year and 5 year,areas under the ROC curves for the development sets were 0.766,0.808 and 0.858 compare with the validation sets were 0.733,0.789 and 0.816 respectively.The calibration curves indicated ideal consistency between the predicted and observed probabilities,and the decision curve analysis presented a clinically useful model.All individuals were allocated into high-risk groups and lowrisk groups based on the median predicted probabilities of the development sets,Kaplan-Meier curve showed there were significant differences in CSS between 2 groups(P<0.001).Conclusion The research established and validated a prognostic nomogram for CSS in Chinese lung adenocarcinoma patients.Application of this model in the clinical setting may assist clinicians in evaluating patient prognosis and providing highly individualized therapy.