肝癌患者术后营养不良影响因素的Lasso-Logistic回归分析及列线图预测模型构建

Lasso-Logistic regression analysis and nomogram prediction model construction of the influencing factors of postoperative malnutrition in patients with hepatocellular carcinoma

目的:采用Lasso-Logistic回归分析肝癌患者术后营养不良的影响因素,并进行列线图预测模型构建,为肝癌患者术后营养状况调整提供参考。方法:选取2021年1月至2023年6月本院收治的460例肝癌手术患者,按照7∶3分为训练队列和验证队列,根据患者主观整体评定量表评判患者术后营养状态,分为营养正常组和营养不良组,采用Lasso-Logistic回归分析其影响因素并构建列线图预测模型。结果:训练队列与验证队列一般临床资料比较差异均无统计学意义(P>0.05)。Lasso-Logistic回归分析显示,年龄、术前体质量指数(BMI)、术前营养风险筛查评分(NRS2002)、贫血、糖尿病、肿瘤分期、白蛋白(ALB)、胰高血糖素样肽-1(GLP-1)、总胆汁酸(TBA)、甘胆酸(CG)均是训练队列中肝癌患者术后营养不良的独立影响因素(P<0.05)。ROC曲线显示,肝癌患者术后营养不良的列线图预测模型在训练队列、验证队列中预测的曲线下面积分别为0.873、0.902;校准曲线显示,该列线图模型在训练队列、验证队列中预测肝癌患者术后营养不良风险与实际风险状况基本一致。结论:构建的列线图模型对肝癌患者术后营养不良预测价值较高,且有良好预测效用,但需要进一步的验证评估。

Objective:To analyze the influencing factors of postoperative malnutrition in patients with liver cancer by Lasso-Logistic regression,and to construct a nomogram prediction model to provide reference for the adjustment of postoperative nutritional status in patients with liver cancer.Methods:A total of 460 patients undergoing surgery for liver cancer in our hospital from January 2021 to June 2023 were selected and divided into a training cohort(322 cases)and a validation cohort(138 cases)according to a ratio of 7∶3.The nutritional status of patients after surgery was evaluated according to the patients'subjective global rating(PG-SGA)scale,and the patients were divided into a normal nutrition group and a malnutrition group.Lasso-Logistic regression was used to analyze the influencing factors of postoperative malnutrition of liver cancer.The nomogram prediction model was constructed by rms package in the training cohort,and the predictive efficiency of the model was verified by receiver operating characteristic(ROC)curve,calibration curve and decision curve(DCA)in the validation cohort.Results:There was no statistically significant difference in general clinical data between the training queue and the validation queue(P>0.05).Lasso-Logistic regression analysis showed that age,preoperative body mass index(BMI),preoperative nutrition risk screening score(NRS2002),anemia,diabetes,tumor stage,albumin(ALB),glucagon like peptide-1(GLP-1),total bile acid(TBA),and glycocholic acid(CG)were all independent influencing factors for postoperative malnutrition of liver cancer patients in the training cohort(P<0.05).The ROC curve showed that the predicted area under the curve of the column chart prediction model for postoperative malnutrition in liver cancer surgery patients in the training and validation queues was 0.873 and 0.902,respectively.The calibration curve showed that the column chart model predicted the risk of postoperative malnutrition in liver cancer surgery patients in the training and validation queues,which was basically consistent with the actual risk status.Conclusion:The constructed nomogram model has high predictive value and good predictive utility for predicting postoperative malnutrition in patients with liver cancer surgery,but further validation and evaluation are needed.