舌鳞状细胞癌患者放疗期间发生口腔感染的模型预测及验证

Model prediction and validation of oral infection during radiotherapy in patients with tongue squamous cell carcinoma

目的:探讨舌鳞状细胞癌患者放疗期间口腔感染的危险因素,旨在建立并验证一种预测舌鳞状细胞癌患者放疗期间口腔感染发生率的可视化评价工具。方法:收集2003~2022年于新疆医科大学第一附属医院就诊的431例放疗的舌鳞癌患者,按2:1的比例随机分为模型组(n=288)和验证组(n=143)。收集患者一般资料。通过单因素及多因素Logistic回归分析进一步探讨并建立列线图预测模型,并由验证组评估舌鳞癌患者放疗期间口腔感染列线图预测模型的可行性。用受试者工作特征(ROC)曲线下面积(AUC)、校准曲线和决策曲线(DCA)分析模型的鉴别能力、准确性和临床实用性。结果:模型组288例放疗舌鳞癌患者中无感染者246例,感染者42例。单因素结合多因素Logistic回归分析结果显示,年龄≥60岁,肿瘤分期Ⅲ~Ⅳ期,口腔环境差,未手术,血清白蛋白低,血红蛋白低是舌鳞癌患者放疗期间口腔感染的独立危险因素(P<0.05)。DCA显示当患者阈值概率为0~0.7,使用列线图预测模型预测舌鳞癌患者放疗期间口腔感染风险的净收益更高。结论:年龄,肿瘤分期,口腔环境,手术,血红蛋白,血清白蛋白是舌鳞癌患者放疗期间发生口腔感染的独立危险因素,基于预测因素建立的列线图模型具有较好的预测效能,有助于个体化评价舌鳞癌患者放疗期间发生口腔感染风险,及时做好防御工作。

Objective:To investigate the risk factors of oral infection during radiotherapy in tongue squamous cell carcinoma patients,with the aim of establishing and validating a visual evaluation tool to predict the incidence of oral infection during radiotherapy in tongue squamous cell carcinoma patients.Methods:Totally 431 tongue squamous cell carcinoma patients who attended the First Affiliated Hospital of Xinjiang Medical University for radiotherapy from 2003 to 2022 were collected,and the patients were randomly divided into model group(n=288)and validation group(n=143)according to a 2∶1 ratio.General information of the patients was collected.The feasibility of the predictive model for oral infection during radiotherapy in tongue squamous cell carcinoma patients was further explored and established by one-way and multi-way logistic regression analysis,and the validation group assessed the feasibility of the predictive model for oral infection during radiotherapy.The discriminatory ability,accuracy and clinical usefulness of the model were analyzed using subject operating characteristic(ROC)area under the curve(AUC),calibration curve and decision curve(DCA).Results:Of 288 patients with radiotherapy tongue squamous cell carcinoma in the model group,246 were free of infection and 42 were infected.Univariate combined with multifactorial Logistic regression analysis showed that age≥60 years,tumor stage Ⅲ~Ⅳ,poor oral environment,no surgery,low serum albumin,and low hemoglobin were independent risk factors for oral infection during radiotherapy in patients with t tongue squamous cell carcinoma(P<0.05).DCA showed that when the threshold probability of patients was 0~0.7,the prediction model using column line graph predicted that patients with tongue squamous cell carcinoma,the net benefit of predicting the risk of oral infection during radiotherapy was higher.Conclusion:Age,tumor stage,oral environment,surgery,he-moglobin,and serum albumin are independent risk factors for oral infection during radiotherapy in tongue squamous cell carcinoma patients,and the column line graph model based on the predictors has better predictive efficacy,which is helpful for individualized evaluation of the risk of oral infection during radiotherapy in tongue squamous cell carcinoma patients and timely defense.