原发性肝癌患者生存期的危险因素及贝叶斯网络模型分析  

Analysis of risk factors and Bayesian network model for the survival of patients with primary hepatic cancer

作　　者：刘赛男 赵和平[2] 韦鳗真 黄河[3] Liu Sainan;Zhao Heping;Wei Manzhen;Huang He(Academy of Medical Sciences,Shanxi Medical University,Taiyuan 030001,Shanxi,China;Department of Gastroenterology,the First Hospital of Shanxi Medical University,Taiyuan 030001,Shanxi,China;Department of General Surgery,First Hospital of Shanxi Medical University,Taiyuan 030001,Shanxi,China)

机构地区：[1]山西医科大学医学科学院,山西太原030001 [2]山西医科大学第一医院消化内科,山西太原030001 [3]山西医科大学第一医院胃肠外科,山西太原030001

出　　处：《肿瘤代谢与营养电子杂志》2024年第5期683-690,共8页Electronic Journal of Metabolism and Nutrition of Cancer

基　　金：山西省卫生健康委科研课题计划项目书(2021052)。

摘　　要：目的探讨影响原发性肝癌(PHC)患者30个月生存期的危险因素及因素间复杂的网络关系。方法回顾性分析2016年10月至2020年11月山西医科大学第一医院363例PHC患者临床资料。通过随访了解患者的生存状况等,从最初诊断时开始计算,以月为单位,观察期截断点为2022年10月。根据文献报道及本研究中PHC患者的平均生存时间(29.6个月),最终将生存时间的研究节点定为30个月。采用多因素Logistic回归分析影响PHC患者30个月生存期的危险因素,并建立贝叶斯网络模型,以深入分析这些危险因素间复杂的网络关系,再通过受试者操作特征(ROC)曲线和校准度散点图(calibration plot)对模型的效能进行评价。结果363例PHC患者中生存时间<30个月的患者有167例(46.01%)。单因素分析结果显示,两组患者在γ-谷氨酰转移酶(γ-GT)水平、肿瘤大小、有无远处转移等方面差异有统计学意义(P<0.05);多因素logistic回归分析结果显示,γ-GT(>40 U/L)、肿瘤大小(>3 cm)、伴有远处转移均为PHC患者30个月生存期的危险因素(P<0.05);贝叶斯网络模型结果显示肿瘤大小、γ-GT水平和有无远处转移是直接影响PHC患者30个月生存期的危险因素,肿瘤个数和年龄则是间接危险因素;条件概率分布表显示当节点肿瘤大小>3 cm、γ-GT>40 U/L,伴有远处转移的患者,PHC生存时间<30个月的概率最大(98.408%)。结论肿瘤大小>3 cm、γ-GT>40 U/L和伴有远处转移是影响PHC患者30个月生存期的危险因素,年龄通过影响肿瘤大小和有无远处转移与PHC患者30个月生存期间接相关。Objective The complex network relationship among risk factors of primary hepatic cancer(PHC)was deeply studied by Bayesian network model,and investigate the risk factors that affect the 30-month survival.Method The clinical data of 363 patients with PHC from October 2016 to November 2020 in the First Hospital of Shanxi Medical University were retrospectively analyzed.Through follow-up surveys,we have learned about the PHC patients'survival status etc.The survival was counted in months,starting from the time of initial diagnosis,and the follow-up surveys ended in October 2022.Based on literature reports and in combination with the average survival of PHC patients in this study(29.6 months),a final decision was made to select 30 months as the research node for survival.The related factors affecting the 30-month survival of PHC patients were analyzed using both univariate and multivariate logistic regression.Additionally,a Bayesian network model was established to further explore the complex network relationship among the risk factors for survival in PHC patients,and to calculate the conditional probability among the factors.The effectiveness of the model was verified by receiver operating characteristic and calibration plot.Result Among the 363 patients with PHC,167 patients had a survival of<30 months(46.01%).The results of the univariate analysis showed significant statistical differences between the two groups of patients in terms ofγ-glutamyl transferase(γ-GT)levels,tumor size,and the occurrence or absence of distant metastasis.The results of the multivariate logistic regression analysis revealed thatγ-GT(>40 U/L),tumor size(>3 cm),and distant metastasis were risk factors for 30-month survival node of PHC patients(P<0.05 for all).The Bayesian network model showed that,for PHC patients,tumor size,γ-GT levels,and the occurrence or absence of distant metastasis were direct risk factors that affected their 30-month survival,while tumor count and age were indirect risk factors that also influenced their 30-month

关 键 词：原发性肝癌 生存时间 危险因素 贝叶斯网络模型 条件概率 预后 相关性 生存率 

分 类 号：R735.7[医药卫生—肿瘤]