非转移性原发性骨肉瘤生存预测模型建立及验证——基于SEER数据库的研究  

Development and validation of a survival prediction model for non-metastatic primary osteosarcoma-a study based on the SEER database

作　　者：曹正一 艾则孜·艾海提[1] 王杰 王志刚[1] CAO Zheng-yi;Aizeizi·Aihaiti;WANG Jie;WANG Zhi-gang(Bone and Joint Movement Ward,Xinjiang Uygur Autonomous Region People’s Hospital,Urumqi,Xinjiang,830001,China)

机构地区：[1]新疆维吾尔自治区人民医院骨关节外科与运动病区,乌鲁木齐830001

出　　处：《中国骨与关节杂志》2025年第2期120-128,共9页Chinese Journal of Bone and Joint

摘　　要：目的基于SEER数据库分析非转移性原发性骨肉瘤预后相关因素,建立并验证非转移性原发性骨肉瘤患者生存预测模型。方法回顾分析2000年至2019年,SEER数据库中病理诊断为骨肉瘤且原发性、非转移性的患者资料。按时间顺序将总体分为2004年至2009年为建模组、2010年至2015年为验证组,使用交叉验证(least absolute shrinkage and selection operator,LASSO)回归筛选变量,将筛选出的变量纳入多因素COX回归分析,确定与结局事件独立相关的因素,建立并验证非转移性原发性骨肉瘤生存预测模型。根据列线图计算患者总得分,使用X-tile软件区分最佳截断值,把总体分为低、中、高风险3组,进行K-M生存分析。结果40~59岁vs.年龄≤19岁组,HR 2.076,95%CI(1.467~2.983)(P<0.001)、≥60岁vs.年龄≤19岁组,HR 2.996,95%CI(1.904~4.731)(P<0.001)。GradeⅢvs.GradeⅠ,HR 3.514,95%CI(1.514~8.153)(P=0.003)、GradeⅣvs.GradeⅠ,HR 2.914,95%CI(1.272~6.675)(P=0.011)。分期Regional vs.Localized,HR 1.603,95%CI(1.198~2.145)(P=0.001)。非手术组vs.手术组,HR 2.166,95%CI(1.293~3.629)(P=0.003)。确定与生存预后独立相关的变量为年龄、分级、分期及手术并构建nomogram列线图,计算建模组N次K折交叉验证时间-AUC值及验证组的时间-AUC值,75%结果>0.7,评估模型具有良好的区分度,绘制建模及验证组1年、3年、5年的校准曲线,提示预测与实际具有较高的一致性。计算出每例患者的nomogram总分,使用X-tile软件进行最佳截断值的划分,将总体分为低风险组:0~121分,中风险组:130~135分,高风险组:139~256分,3组K-M分析显示生存差异有统计学意义,P<0.0001。结论利用SEER数据库对非转移性原发性骨肉瘤患者进行生存预后分析,建立了1年、3年、5年的生存预测模型,使用该模型有助于临床计算生存率,以便加强管理,进一步改善疾病的预后。Objective To analyze the prognostic factors related to non-metastatic primary osteosarcoma based on the SEER database,and to establish and validate a survival prediction model for patients with non-metastatic primary osteosarcoma.Methods Retrospectively analyze the data of patients with pathological diagnosis of osteosarcoma and primary,non-metastatic in SEER database from 2000-2019,exclude some patients by exclusion criteria,define the outcome event as all-cause death,and chronologically divide the overall into the modeling group in 2004-2009 and the validation group in 2010-2015.The cross-validation LASSO(least absolute shrinkage and selection operator)regression was used to screen variables,and the screened variables were included in a multifactor COX regression analysis to identify factors independently associated with the outcome event,and to establish and validate a survival prediction model for non-metastatic primary osteosarcoma.Total patient scores were calculated based on columnline plots,and the best cutoff values were differentiated using X-tile software to classify the overall population into low,medium,and high risk groups for K-M survival analysis.Results The age:age 40-59 years vs.age≤19 years,HR 2.076,95%CI(1.467-2.983),(P<0.001);age≥60 years old vs.age≤19 years,HR 2.996,95%CI(1.904-4.731),(P<0.001).GradeⅢvs.GradeⅠ,HR 3.514,95%CI(1.514-8.153),(P=0.003).GradeⅣvs.GradeⅠ,HR 2.914,95%CI(1.272-6.675),(P=0.011).Staging-regional vs.localized,HR 1.603,95%CI(1.198-2.145),(P=0.001).Non-surgical vs.surgical,HR 2.166,95%CI(1.293-3.629),(P=0.003).Variables independently associated with survival prognosis were determined as the age,grade,stage,and surgery,and nomogram was constructed to calculate the time-AUC values for N-fold k-fold cross-validation for the modeling group and the time-AUC values for the validation group.The result as 75%was greater than 0.7,which suggested that the model had good differentiation.The 1,3,5-year calibration curves of the modeling and validation groups were plotted

关 键 词：骨肉瘤 模型 预测 预后 列线图 原发性 非转移性 

分 类 号：R738.1[医药卫生—肿瘤]