个体化预测非小细胞肺癌患者化疗期间肺部感染风险Nomogram模型的建立与验证

Construction and validation of Nomogram model for individualized prediction of risk of pulmonary infection in patients with non-small cell lung cancer during chemotherapy

目的构建个体化预测非小细胞肺癌(NSCLC)患者化学治疗(化疗)期间发生肺部感染的风险Nomogram模型,并对模型的预测效能进行验证。方法选取2018年2月—2021年1月在某院进行治疗的218例NSCLC患者作为研究对象,根据化疗后是否发生肺部感染将其分为肺部感染组(n=56)和非肺部感染组(n=162),采用LASSO和logistic回归分析筛选NSCLC患者化疗期间发生肺部感染的独立危险因素,并构建Nomogram预测模型。结果纳入218例NSCLC患者,56例化疗期间发生肺部感,感染发病率为25.69%。LASSO和logistic回归分析结果显示:年龄≥60岁、糖尿病、联合化疗药物、化疗周期>2次、化疗后清蛋白含量<30 g/L、化疗前KPS评分<80分是NSCLC患者化疗期间发生肺部感染的独立预测因素(均P<0.05)。基于6项独立预测因素建立预测NSCLC患者化疗期间发生肺部感染的风险Nomogram模型,验证结果显示,训练集和验证集的C-index分别为0.819(95%CI:0.788~0.850)、0.802(95%CI:0.778~0.829),两集校准曲线走向与对角线(理想曲线)较为贴近,受试者工作特征曲线下面积分别为0.807(95%CI:0.775~0.839)、0.797(95%CI:0.773~0.821),决策曲线显示阈值概率在1%~90%时,有较高的净获益值。结论基于NSCLC患者化疗期间发生肺部感染独立预测因素建立的Nomogram模型,具有较好的预测效能,有助于临床及早筛查高风险患者和进一步改进治疗计划。

Objective To construct a Nomogram model for individualized prediction of risk of pulmonary infection(PI)in patients with non-small cell lung cancer(NSCLC)during chemotherapy period,and validate the prediction efficiency of the model.Methods 218 patients with NSCLC who were treated in a hospital from February 2018 to January 2021 were selected as the research objects,according to whether they had PI after chemotherapy,they were divided into PI group(n=56)and pulmonary non-infection group(n=162),LASSO and logistic regression analysis were used to screen the independent risk factors for PI in NSCLC patients during chemotherapy,and a Nomogram prediction model was constructed.Results 218 NSCLC patients were included in study,56 patients had PI during chemotherapy,incidence of PI was 25.69%.LASSO and logistic regression analysis showed that age≥60 years old,diabetes mellitus,combined chemotherapy drugs,chemotherapy cycle>2 times,albumin content after chemotherapy<30 g/L,KPS score<80 points before chemotherapy were independent predictors of PI in NSCLC patients during chemotherapy(all P<0.05).Based on 6 independent predictors,the Nomogram model was constructed to predict the risk of PI in NSCLC patients during chemotherapy.Validation result showed that the C-index of the training set and validation set were 0.819(95%CI:0.788-0.850)and 0.802(95%CI:0.778-0.829)respectively.The calibration curve trend of two sets was relatively close to the diagonal(ideal curve),the area under the ROC curve(AUC)were 0.807(95%CI:0.775-0.839)and 0.797(95%CI:0.773-0.821)respectively.When the decision curve showed that the threshold probability was 1%-90%,there was a relatively high net profit value.Conclusion Nomogram model based on independent predictors of PI in NSCLC patients du-ring chemotherapy has good prediction efficiency,which is helpful to screen high-risk patients as early as possible and improvement of treatment plans.