基于预后营养指数的鼻咽癌患者2级及以上放射性口腔黏膜炎预测模型

文献信息

中文摘要

引言： 构建基于预后营养指数的鼻咽癌患者2级及以上放射性口腔黏膜炎预测模型，为早期识别高危患者提供依据。
方法： 回顾性纳入2020年5月至2025年12月在蚌埠医科大学第一附属医院接受放化疗的192例原发性鼻咽癌患者。患者按7:3比例随机分配至训练集（n=133）和内部验证集（n=59）。在正态性和方差齐性检验后比较基线特征。联合使用LASSO、随机森林和支持向量机递归特征消除筛选候选变量，随后采用单因素和多因素logistic回归识别与2级及以上放射性口腔黏膜炎相关的因素。构建列线图，并使用Firth惩罚似然回归以减少小样本偏倚。采用受试者工作特征分析和校准曲线评估模型性能，通过决策曲线分析评价临床实用性。随后使用倾向性评分匹配进行敏感性分析。
结果： 预后营养指数是2级及以上放射性口腔黏膜炎的独立保护因素。训练集曲线下面积为0.842，验证集为0.725。预后营养指数高值组患者风险显著较低（OR=0.067，P<0.001），且该发现在Firth回归分析中保持一致。在1:1倾向性评分匹配后，该关联仍显著（OR=0.709，P<0.001；AUC=0.810）。
结论： 预后营养指数是2级及以上放射性口腔黏膜炎的独立预测因子，该模型对接受放化疗的鼻咽癌患者这一结局显示出有用的预测性能。

英文摘要

INTRODUCTION: To develop a prediction model based on the prognostic nutritional index for grade 2 or higher radiation-induced oral mucositis in patients with nasopharyngeal carcinoma, providing a basis for early identification of patients at high risk.
METHODS: We retrospectively included 192 patients with primary nasopharyngeal carcinoma who received chemoradiotherapy at the First Affiliated Hospital of Bengbu Medical University between May 2020 and December 2025. Patients were randomly assigned in a 7:3 ratio to a training set (n = 133) and an internal validation set (n = 59). Baseline characteristics were compared after tests of normality and homogeneity of variance. Candidate variables were selected jointly using LASSO, random forest, and support vector machine recursive feature elimination, followed by univariable and multivariable logistic regression to identify factors associated with grade 2 or higher radiation-induced oral mucositis. A nomogram was constructed, and Firth penalized likelihood regression was used to reduce small-sample bias. Model performance was assessed using receiver operating characteristic analysis and calibration curves, while clinical utility was evaluated by decision curve analysis. Propensity score matching was then used for sensitivity analysis.
RESULTS: The prognostic nutritional index was an independent protective factor of grade 2 or higher radiation-induced oral mucositis. The area under the curve was 0.842 in the training set and 0.725 in the validation set. Patients in the group with high prognostic nutritional index values had a markedly lower risk (OR = 0.067, P < 0.001), and the finding remained consistent in the Firth regression analysis. The association also remained significant after 1:1 propensity score matching (OR = 0.709, P < 0.001; AUC = 0.810).
CONCLUSION: The prognostic nutritional index is an independent predictor of grade 2 or higher radiation-induced oral mucositis, and the model showed useful predictive performance for this outcome in patients with NPC receiving chemoradiotherapy.