【摘要】 目的 利用Dryad数据库构建基于血浆免疫标志物预测听神经瘤（VS）患病风险的模型。方法 对Dryad数据库中的患者资料进行二次分析，按是否罹患VS分为VS组（213例）和非VS组（113例），采用单因素分析、多因素Logistic回归及LASSO回归筛选危险因素并构建列线图预测模型。结果 单因素分析显示，两组在血浆标志物肿瘤坏死因子受体2(TNF-R2)、巨噬细胞迁移抑制因子（MIF）、基质细胞衍生因子-1α(SDF-1α)、白细胞介素-2受体（IL-2R）、单核细胞趋化蛋白-2(MCP-2)、CD30分子（CD30）、白细胞介素-16(IL-16)、肿瘤坏死因子样弱凋亡诱导因子（TWEAK）、B淋巴细胞趋化因子（BLC）、单核细胞趋化蛋白-3(MCP-3)等的表达水平比较，差异均有统计学意义（P<0.05）。经LASSO回归分析筛选表明，SDF-1α[OR(95%CI)=1.001(1.000～1.001)]、IL-2R[OR=1.002(1.001～1.003)]、BLC[OR=1.015(1.006～1.023)]、MCP-3[OR=1.294(1.159～1.444)]为VS患病的独立危险因素（P<0.05）。以此4个变量构建列线图模型，受试者工作特征（ROC）曲线分析显示模型曲线下面积（AUC）为0.856(95%CI:0.815～0.897)，提示预测效能良好。Bootstrap法重抽样1 000次进行模型内部验证，校准曲线与理想曲线接近，表明预测准确性较高。决策曲线分析（DCA）显示，当阈值概率为0～60%时，该模型的净收益显著高于“全干预”与“无干预”策略，具有临床实用价值。结论 基于LASSO回归筛选血浆免疫标志物所构建的列线图模型可用于预测VS患病风险，为临床早期诊断提供参考依据。更多还原

【Abstract】 Objective To construct a model for predicting the risk of vestibular schwannoma(VS) using plasma immune markers based on the Dryad database. Methods A secondary analysis was performed on patient data from the Dryad database. Participants were divided into a VS group(n=213) and a non-VS group(n=113). Univariate analysis, multivariate Logistic regression, and LASSO regression were used to screen risk factors and construct a nomogram prediction model. Results Univariate analysis revealed statistically significant differences(P<0.05) in plasma markers including TNF-R2, MIF, SDF-1α, IL-2R, MCP-2, CD30, IL-16, TWEAK, BLC, and MCP-3 between the two groups. Multivariate Logistic regression combined with LASSO regression identified SDF-1α [OR(95%CI) = 1.001(1.000～1.001)], IL-2R [OR(95%CI)=1.002(1.001～1.003)], BLC [OR(95%CI)=1.015(1.006-1.023)], and MCP-3 [OR(95%CI)=1.294(1.159-1.444)] as independent risk factors for VS. A nomogram model was constructed using these four variables. The analysis of the receiver operating characteristic(ROC) curve revealed an area under the curve(AUC) value of 0.856(95% CI: 0.815-0.897), which suggested a high level of predictive accuracy. The model exhibited strong predictive accuracy, as evidenced by a well-calibrated curve that closely aligned with the ideal line, achieved through bootstrap resampling with 1 000 iterations. Furthermore, decision curve analysis(DCA) indicated robust clinical utility, with the decision curve surpassing both the "None" and "All" reference lines across a threshold probability range of 0-60%. Conclusions The nomogram prediction model for VS risk, constructed on plasma immune markers screened by LASSO regression, can provide a reference for the early clinical diagnosis.更多还原