生存树方法在肿瘤预后分析中的应用  

An application of survival tree method in the tumor prognostic analysis

作　　者：郎素平[1] 余红梅[2] 王彤[2] 何大卫[2] 

机构地区：[1]南京医科大学公共卫生学院流行病与卫生统计学系,江苏南京210029 [2]山西医科大学公共卫生学院卫生统计学教研室,山西太原030001

出　　处：《疾病控制杂志》2005年第6期557-560,共4页Chinese Journal of Disease Control and Prevention

摘　　要：目的对于肿瘤病人的预后分析,传统方法多集中于对预后相关因素的探讨,而由生存树方法得到的预后分组不仅可以有助了解具有相似预后人群的临床特征,还可以从中发现传统的生存模型不易发现的交互作用。方法本文结合生存树方法与传统的Cox回归模型,对235例胃癌病人进行预后分析。结果在Cox回归中,淋巴结转移、肿瘤大小、手术切缘有无癌组织作为3个独立的预后因素被筛选出来;对该资料进行生存树分析,得到3个预后子群,其中位生存期分别为24个月、12个月、5个月。结论将生存树方法与Cox回归模型相结合,可以得到更完善的预后分析结论。Objective The traditional model focused on assessing the relative prognostic factors, while the survival tree method could identify subsets of patients with homogeneous clinical feature. It was also useful for detecting nonlinear interactions between baseline variables. Methods The survival tree and Cox regression were applied to analyze prognostic among 235 patients with gastric cancer. Results Lymph nodemetastasis, tumor size and cancer cells of operation cutting were selected to be independent factors in Cox regression, three subgroups of patients were found with median survival times of 24, 12 and 5 months respectively. Conclusions Combined with Cox regression, the survival tree method may be helpful to perfect the prognostic analysis,

关 键 词：存活率 比例危险度模型 预后 

分 类 号：R195.1[医药卫生—卫生统计学] R818.07[医药卫生—卫生事业管理]