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 of γ-glutamyl transferase γ-GT levels tumor size and the occurrence or absence of distant metastasis. The results of the multivariate logistic regression analysis revealed that γ-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 survival. The conditional probability distribution table showed that tumor size >3 cm γ-GT >40 U/ L and distant metastasis in patients with PHC had the highest probability of survival <30 months 98. 408% . Conclusion The tumor size >3 cm γ-GT level >40 U/ L and distant metastasis are the risk factors affecting the 30-month survival of PHC patients. Age have a indirect relationship with the 30-month survival of PHC patients by affecting tumor size and the occurrence or absence of distant metastasis.