摘要
目的:为减少甲状腺癌诊疗领域存在的过度诊断现象,开发一种基于体检数据进行甲状腺异常增殖风险的数学预测模型。方法:研究数据来源于福建泉州中国人民解放军第910医院健康管理中心,分析研究了血清胸苷激酶1(STK1)与甲状腺影像报告及数据系统(TI-RADS)分级的相关性。以TI-RADS分级作为终点事件,以STK1检测结果联合常规体检指标为自变量,通过不平衡类学习算法(IKIL)构建了甲状腺异常增殖风险的风险预测模型,并绘制受试者工作特征曲线(ROC),以评价此模型在受检者甲状腺异常增殖风险评估中的应用价值。结果:K-S检验结果显示STK1值呈偏态分布。由Kruskal-Wallis检验可知,随着TI-RADS分级数的升高,STK1水平呈现单调递增趋势。从TR-1进展到TR-2时细胞增殖水平显著增加;在TR-2到TR-4区间,细胞增殖水平存在升高趋势,但相邻级别之间差异并不显著。通过IKIL法获得的甲状腺增殖风险指数公式为:y=-3.50+0.29×STK1+0.06×总蛋白+0.01×乳酸脱氢酶-0.0046×尿酸-1.35×高密度脂蛋白胆固醇+0.34×CEA,该模型的ROC验证AUC值为0.848,标准误差0.00925,95%置信区间为(0.827,0.865)。Youden指数最大的增殖风险指数为0.52,使用该点为阈值预测甲状腺异常增殖风险的敏感性、特异性分别为66.98%、94.22%,提示风险预测模型对甲状腺异常增殖风险具有较好地预测能力。结论:研究开发了基于STK1水平联合其他体检指标对受检者的甲状腺异常增殖风险的预测模型。该风险预测模式的限制条件少,可操作性强,对社会自然人群的甲状腺结节/肿瘤的健康管理具有较好现实意义。
Objective:In order to reduce overdiagnosis in thyroid cancer treatment,a risk prediction model of thyroid abnormal proliferation based on physical examination data was developed.Methods:In thisstudy,we analyzed the correlation between serum thymidine kinase 1 level and thyroid imaging report and data system grading by using the data collected from 2014 to 2017 in the 910 Hospital of the Chinese People's Liberation Army in Quanzhou,Fujian province.With TI-RADS grade wastaken asthe end point,serumTK1 detection results combined with physical exam indexes was taken as independent variables,a risk prediction model of thyroid abnormal proliferation was constructed by using imbalanced Learning based on K-means and Logistic Regression algorithm.Furthermore,application value of this risk prediction model was evaluated with operating characteristic curve.Results:STK1 values oberserved skewed distribution by Kolmogorov-Smirmov test.According to Kruskal-Wallis test result,with the increase of the number of TI-RADS grades,the level of STK1 presents a monotonically increasing trend.Cell proliferation was significantly increased fromTR-1 toTR-2.In theTR-2 toTR-4 range,there was an increased trend of cell proliferation,but with no significant difference between adjacent grades.The formula of thyroid proliferation risk index obtained by IKIL method:Y=-3.50+0.29×STK1+0.06×total protein+0.01×lactate dehydrogenase-0.0046×uric acid-1.35×high-density lipoprotein+0.34×CEA,The AUC value verified by ROC of this model was 0.848,standard error was 0.00925,and 95%confidence interval was(0.827,0.865).The maximum proliferation risk of The Youden index is 0.52.The sensitivity and specificity of using this threshold to predict the risk of thyroid abnormal proliferation were 66.98%and 94.22%,respectively.Risk prediction ability of this model of thyroid abnormal proliferation was proved by results above.Conclusion:In this study,we developed a predictive model which can preliminarily assess the risk of thyroid abnormal proliferation according to STK1 level combined with other physical examination indicators.With low requirement and strong operability,this risk prediction model has a good identification significance for the health management of benign and malignant thyroid tumors in the general population.
作者
刘绵学
江小蓉
兰丽云
汪海东
黑爱莲
李劲
周际
王瑜
LIU Mianxue;JIANG Xiaorong;LAN Liyun;WANG Haidong;HEI Ailian;LI Jin;ZHOU Ji;WANG Yu(Shenzhen Ellen-Sven Precision Medicine Institute,Shenzhen 518038,Guangdong,China;Fujian Medicl University 2nd Affiiated Hospital,Fuzhou 350108,Fujian,China;Health Management Center,PLA 910 Hospital,Quanzhou 362000,Fujian,China)
出处
《健康体检与管理》
2022年第3期226-232,共7页
Journal of Health Examination and Management