Logistic回归和ROC曲线分析血清DKK1、GP73和AFP在原发性肝癌诊断中的价值  

Evaluation on the diagnostic significance of DKK1,GP73 and AFP in primary hepatic carcinoma by Logisitic regression analysis and ROC curve analysis

作　　者：宗迎迎 徐浩[1] 许伟[1] 顾玉明[1] 祖茂衡[1] 万豪光[1] 吴康[1] 

机构地区：[1]徐州医学院附属医院介入科,江苏徐州221002

出　　处：《检验医学》2015年第6期559-563,共5页Laboratory Medicine

摘　　要：目的应用Logistic回归和受试者工作特征（ROC）曲线探讨血清Dickkopf同源物1（DKK1）、高尔基体糖蛋白73（GP73）和甲胎蛋白（AFP）对原发性肝癌（PHC）的诊断价值。方法采用酶联免疫吸附试验（ELISA）测定81例PHC、50例慢性肝炎（CH）、50例肝硬化（LC）患者和50名健康对照（HC）者血清DKK1、GP73浓度,采用电化学发光免疫法（ECLIA）测定AFP浓度,建立PHC诊断的统计模型即综合预测模型,应用Logistic回归和ROC曲线进行分析。结果 PHC组DKK1水平[5.29（4.27-7.45）ng/mL]明显高于CH组[4.11（3.82-4.52）ng/mL]、LC组[3.66（3.61-4.15）ng/mL]和HC组[3.42（3.28-3.68）ng/mL]（P〈0.01）。PHC组GP73水平[（215.44±68.66）ng/mL]明显高于CH组[（121.02±49.10）ng/mL]、LC组[（168.39±37.91）ng/mL]和HC组[（57.36±19.94）ng/mL]（P〈0.05）。DKK1、GP73、AFP诊断PHC的敏感性分别为71.6%、77.8%和58.0%,特异性分别为90.0%、86.7%和78.7%。综合预测模型的敏感性和特异性为87.7%和90.0%。AFP、DKK1、GP73的ROC曲线下面积（AUC）分别为0.76、0.86、0.87,综合预测模型AUC为0.93,与AFP、DKK1、GP73比较,差异均有统计学意义（P〈0.01）。结论 DKK1、GP73可以作为诊断PHC的血清标志物,与AFP联合检测可提高诊断效能;Logistic回归结合ROC曲线分析简单、有效,可用于PHC多指标联合诊断的分析评价。Objective To investigate the diagnostic significance of serum Dickkopf-1（ DKK1）,Golgi protein 73（GP73） and alpha fetoprotein（AFP） in primary hepatic carcinoma（PHC） by Logistic regression analysis and receiver operating characteristic（ ROC） curve analysis. Methods Enzyme-linked immunosorbent assay（ ELISA） and electrochemiluminescence immunoassay（ ECLIA） were used to determine the serum levels of DKK1,GP73 and AFP in81 patients with PHC,50 patients with chronic hepatitis（ CH）,50 patients with liver cirrhosis（ LC） and 50 healthy controls（ HC）. A statistical model for diagnosing PHC was established,and the results were analyzed by Logisitic regression analysis and ROC curve analysis. Results Serum DKK1 levels in PHC group [5. 29（4. 27-7. 45） ng / mL]were significantly higher than those in CH group[4. 11（3. 82-4. 52） ng / mL],LC group [3. 66（3. 61-4. 15） ng / mL]and HC group[3. 42（3. 28-3. 68） ng / mL]（ P〈0. 01）. Serum GP73 levels in PHC group[（215. 44 ± 68. 66） ng / mL]were significantly higher than those in CH group [（121. 02 ± 49. 10） ng / mL],LC group[（168. 39 ± 37. 91） ng / mL]and HC group[（57. 36 ± 19. 94） ng / mL]（ P〈0. 05）. The sensitivities and specificities of DKK1,GP73 and AFP were71. 6%,77. 8%,58. 0% and 90. 0%,86. 7%,78. 7%,respectively. The model had a sensitivity of 87. 7% and a specificity of 90. 0%. The areas under ROC curves（ AUC） of AFP,DKK1 and GP73 were 0. 76,0. 86 and 0. 87,respectively,and the AUC of the model was 0. 93,having statistical significance with AFP,DKK1 and GP73（ P〈0. 01）. Conclusion DKK1 and GP73 can be as serum markers for the diagnosis of PHC, and the combined determination of DKK1,GP73 and AFP could improve PHC diagnostic performance. Moreover,Logisitic regression analysis combined with ROC curve analysis is a simple and effective statistical method,and it can be used in the analysis and evaluation of multi-index combined diagnosis of PHC.

关 键 词：Dickkopf同源物1 高尔基体糖蛋白73 甲胎蛋白 原发性肝癌 诊断 

分 类 号：R446.61[医药卫生—诊断学]