原发性肝癌微创切除术后近期预后不良的风险因素及个体化预警模型的构建研究

Risk factors for poor short-term prognosis after minimally invasive resection of primary liver cancer and construction of individualized warning model

目的分析原发性肝癌微创切除术后近期预后不良的风险因素,并构建个体化预警模型。方法回顾性分析2018年1月—2022年5月郑州大学第一附属医院收治的357例行微创切除术的原发性肝癌患者的临床资料,将其按照2∶1比例随机分为建模集(n=238)和验证集(n=119)。根据术后1年内预后不良的发生情况将建模集患者分为预后不良组(n=57)和预后良好组(n=181),采用logistic回归分析筛选原发性肝癌微创切除术后近期预后不良的风险因素,构建个体化预警模型并对该模型进行验证。结果建模集数据logistic回归分析结果显示,肿瘤直径>5cm、低分化、血管侵犯、门静脉浸润、肝硬化、甲胎蛋白(AFP)≥400μg/L是原发性肝癌微创切除术后近期预后不良的危险因素(P<0.05);基于以上6项指标构建原发性肝癌微创切除术后近期预后不良的个体化预警模型,建模集和验证集的校准曲线显示该模型校准曲线与理想曲线一致性良好,一致性指数分别为0.816(95%CI:0.763~0.862)、0.805(95%CI:0.724~0.870);受试者工作特征曲线显示该模型预测建模集和验证集患者术后近期预后不良的曲线下面积分别为0.823(95%CI:0.768~0.869)、0.810(95%CI:0.727~0.876);决策曲线显示,建模集阈概率在0~1.0时均可获得净收益,验证集阈概率在0~0.81时可获得净收益。结论肿瘤直径>5cm、低分化、血管侵犯、门静脉浸润、肝硬化、AFP≥400μg/L是原发性肝癌微创切除术后近期预后不良的危险因素,以此构建的个体化预警模型的预测效能和适用性良好,可用于预测原发性肝癌患者预后。

Objective To analyze the risk factors for poor short-term prognosis after minimally invasive resection of primary liver cancer,and construct an individualized warning model.Methods The clinical data of 357 patients with primary liver cancer undergoing minimally invasive resection admitted to the First Affiliated Hospital of Zhengzhou University from January 2018 to May 2022 were retrospectively analyzed,and they were divided into modeling set(n=238)and validation set(n=119)according to the ratio of 2∶1.The modeling set patients were divided into poor prognosis group(n=57)and good prognosis group(n=181)based on the occurrence of poor prognosis within 1 year after surgery.The risk factors for poor short-term prognosis after minimally invasive resection of primary liver cancer was screened through logistic regression analysis,and an individualized warning model was constructed and validated.Results Logistic regression analysis results of the modeling set data showed that tumor diameter>5cm,poorly differentiated,vascular invasion,portal vein infiltration,liver cirrhosis and alpha fetoprotein(AFP)≥400μg/L were the risk factors for poor short-term prognosis after minimally invasive resection of primary liver cancer(P<0.05).Based on the above 6 indicators,the individualized warning model for poor short-term prognosis after minimally invasive resection of primary liver cancer was established.The calibration curves of the modeling set and validation set showed that the calibration curve of this model had good consistency with the ideal curve,and the consistency index of them were 0.816(95%CI:0.763~0.862)and 0.805(95%CI:0.724~0.870)respectively.The receiver operating characteristic showed that the area under the curve predicted by this model for poor short-term prognosis after surgery in the modeling set and validation set were 0.823(95%CI:0.768~0.869)and 0.810(95%CI:0.727~0.876)respectively.The decision curve showed that the modeling set could obtain net benefits when the threshold probability was 0-1.0,while the validation set could obtain net benefits when the threshold probability was between 0-0.81.Conclusion Tumor diameter>5 cm,poorly differentiated,vascular invasion,portal vein infiltration,liver cirrhosis and alpha fetoprotein(AFP)≥400μg/L are risk factors for poor short-term prognosis after minimally invasive resection of primary liver cancer,and the individualized warning model constructed based on them has good predictive efficiency and applicability,and it can be used to predict the prognosis of patients with primary liver cancer.