多模态代谢-血液标志物联合甲状腺超声特征构建的高危结节风险预测列线图

Nomogram for predicting high-risk thyroid nodules based on multimodal metabolic and blood markers combined with ultrasound characteristics

摘要:

目的 探讨多模态代谢-血液标志物联合甲状腺超声特征在甲状腺高低危结节鉴别中的价值，并构建高危结节风险预测列线图模型。方法 选取2024年5月—2025年4月淮安市第一人民医院健康管理中心检出的8 141例甲状腺结节患者，根据甲状腺超声甲状腺影像报告和数据系统分类标准，将4类结节（4a、4b、4c）归为高危结节组（111例），1～3类结节归为低危结节组（8 030例）。比较两组多模态代谢-血液标志物水平及甲状腺结节超声特征差异，通过多因素一般Logistic回归模型分析影响高危结节的独立危险因素，并构建甲状腺高危结节发生的预测列线图模型及校正曲线。结果 高危结节组年龄、促甲状腺激素（TSH）、低密度脂蛋白胆固醇（LDL-C）水平均高于低危结节组（P <0.05），游离甲状腺素（FT4）水平低于低危结节组（P <0.05）。低危结节组血流分级0、Ⅰ级占比、边界清晰率、形态规则率均高于高危结节组（P <0.05），钙化率和内部低回声占比均低于高危结节组（P <0.05）。多因素一般Logistic回归分析结果显示：年龄大［O^R=1.036（95% CI：1.011，1.061）］、TSH水平高［O^R=1.058（95% CI：1.017，1.100）］、LDL-C水平高［O^R=1.644（95% CI：1.065，2.537）］、发生钙化［O^R=12.767（95% CI：6.583，24.762）］、血流分级Ⅱ、Ⅲ级［O^R=10.351（95% CI：5.347，20.038）］、内部回声低［O^R=17.075（95% CI：8.343，34.943）］均为发生甲状腺高危结节的危险因素（P <0.05）；FT4水平高［O^R=0.826（95% CI：0.700，0.973）］、边界清晰［O^R=0.046（95% CI：0.023，0.091）］、形态规则［O^R=0.019（95% CI：0.009，0.039）］均为甲状腺高危结节的保护因素（P <0.05）；并构建甲状腺高危结节发生的列线图模型。结论 基于多模态代谢-血液标志物与甲状腺超声特征构建的列线图模型能有效区分甲状腺高低危结节，为临床决策提供量化工具。

Abstract:

Objective To investigate the value of multimodal metabolic and blood biomarkers combined with thyroid ultrasound characteristics in distinguishing benign low-risk from high-risk thyroid nodules, and to construct a risk prediction nomogram model for high-risk nodules.Methods A total of 8 141 patients with thyroid nodules detected at the Health Management Center of Huai'an First People's Hospital from May 2024 to April 2025 were enrolled. According to the Thyroid Imaging Reporting and Data System (TI-RADS) classification criteria, category 4 nodules (4a, 4b, and 4c) were defined as the high-risk group (n = 111), while categories 1-3 nodules were defined as the low-risk group (n = 8 030). Differences in levels of multimodal metabolic and blood markers and ultrasound characteristics of thyroid nodules were compared between the groups. Independent risk factors for high-risk nodules were analyzed using a multivariable logistic regression model, and a nomogram along with a calibration curve was constructed to predict the occurrence of high-risk thyroid nodules.Results The high-risk group exhibited higher age and levels of TSH and LDL-C (P < 0.05), while FT4 levels were lower in the high-risk group than in the low-risk group (P < 0.05). The low-risk group had higher proportions of nodules with grades 0 and I blood flow, clear margins, and regular morphology (P < 0.05), alongside lower rates of nodules with calcification and internal hypoechogenicity compared to the high-risk group (P < 0.05). Multivariable logistic regression analysis revealed that advanced age [O^R = 1.036 (95% CI: 1.011, 1.061) ], elevated TSH levels [O^R = 1.058 (95% CI: 1.017, 1.100) ], high LDL-C levels [O^R = 1.644 (95% CI: 1.065, 2.537) ], calcification [O^R = 12.767 (95% CI: 6.583, 24.762) ], grade Ⅱ~Ⅲ blood flow [O^R = 10.351 (95% CI: 5.347, 20.038) ], and internal hypoechogenicity [O^R = 17.075 (95% CI: 8.343, 34.943) ] were risk factors for high-risk thyroid nodules (P < 0.05). Conversely, elevated FT4 levels [O^R = 0.826 (95% CI: 0.700, 0.973) ], well-defined borders [O^R = 0.046 (95% CI: 0.023, 0.091) ], and regular morphology [O^R = 0.019 (95% CI: 0.009, 0.039) ] were protective factors against high-risk thyroid nodules (P < 0.05). A nomogram model for predicting high-risk thyroid nodules was constructed.Conclusion The nomogram model, constructed based on multimodal metabolic and blood biomarkers and thyroid ultrasound characteristics, effectively distinguishes between high-risk and low-risk thyroid nodules, providing a quantitative tool for clinical decision-making.