【摘要】 目的 研究初诊2型糖尿病（T2DM）患者群体中，成人隐匿性自身免疫糖尿病（LADA）的检出率及谷氨酸脱羧酶抗体（GADA）在鉴别LADA和T2DM临床应用中的价值。方法 选取2022年1月至2024年1月江苏省中医院检验科收集的113例初诊T2DM患者为研究对象，检测其血清中的GADA，根据诊疗标准将患者划分为LADA组和T2DM组，并对这两组进行临床资料和生化检测结果的对比研究。为了评价相关指标在区分LADA和T2DM的诊断价值，绘制受试者工作特征（ROC）曲线；运用决策树机器学习模型，进一步评价指标并区分LADA和T2DM的效果，在训练后初步预测诊断。结果 初诊为T2DM患者中GADA>10 IU/ml的患者中，GADA在区分LADA和T2DM患者方面具有非常好的区分度，AUC值高达0.801；空腹C肽（CP）在区分LADA和T2DM患者方面具有较好的区分度，AUC值为0.651；糖化血红蛋白（HbA1c）在区分LADA和T2DM患者方面的区分度有限，AUC值为0.374，但仍具有一定的统计学意义。决策树模型不仅验证了描述性、相关性和ROC曲线分析的结论，且在少样本的条件下，利用GADA、CP和HbA1c指标实现诊断预测精度达0.760。结论 对初诊为T2DM的患者进行GADA检测是筛查LADA的重要指标，而CP和HbA1c作为次要指标，有助于提高LADA的诊断准确性。本研究的结果支持在临床实践中应用GADA检测，并结合决策树模型辅助诊断LADA，为LADA早期诊断和治疗提供了新的工具。更多还原

【Abstract】 Objective To investigate the detection rate of latent autoimmune diabetes in adults(LADA) in newly diagnosed patients with type 2 diabetes mellitus(T2DM) and the value of glutamic acid decarboxylase antibody(GADA) in differentiating LADA from T2DM in clinical application. Methods A total of 113 newly diagnosed patients with T2DM admitted to and collected by the Department of Clinical Laboratory of Jiangsu Province Hospital of Chinese Medicine from January 2022 to January 2024 were selected as the research objects. GADA in their serum was detected, and the patients were divided into the LADA group and the T2DM group according to the diagnosis and treatment criteria. Clinical data and biochemical test results were compared between these two groups. To evaluate the diagnostic value of relevant indices in distinguishing LADA from T2DM, receiver operating characteristic(ROC) curves were plotted. Using decision tree machine learning models, indices were further evaluated and the effectiveness of LADA and T2DM were distinguished, and diagnosis was preliminarily predicted after training. Results In patients with initial diagnosis of T2DM and GADA>10 IU/ml, GADA had a very good discrimination ability between LADA and T2DM patients, with an AUC value of up to 0.801. C-peptide(CP) had good discrimination between LADA and T2DM patients, with an AUC value of 0.651. The discriminatory power of glycated hemoglobin(HbA1c) in distinguishing LADA and T2DM patients was limited, with an AUC value of 0.374, but it still had some statistical significance. The decision tree model not only validated the conclusions of descriptive, correlation, and ROC curve analysis, but also achieved a diagnostic prediction accuracy of 0.760 using GADA,CP, and HbA1c indices under small sample conditions. Conclusion GADA testing is an important index for screening LADA in patients with newly diagnosed T2DM, while CP and HbA1c are secondary indices that help improve the diagnostic accuracy of LADA. The results of this research support the application of GADA detection in clinical practice, combined with decision tree model to assist in the diagnosis of LADA, providing a new tool for early diagnosis and treatment of LADA.更多还原