自主研发CEA、ProGRP检测系统与国际同类产品的结果一致性分析和临床诊断效能评价

Independent Development of CEA and ProGRP Detection Systems: Analysis of Result Consistency with International Counterparts and Evaluation of Clinical Diagnostic Performance

摘要: 目的：系统评估国产自主研发癌胚抗原(CEA)、胃泌素释放肽前体(ProGRP)检测系统方法学性能，及其在消化道肿瘤早期诊断的应用价值，并与雅培、罗氏两大国际品牌检测系统开展平行对比，明确该检测系统在真实临床场景中的应用潜力。方法：本检测系统参考CLSI EP9-A2和EP12-A2文件，采用宁波海尔施智造自主研发检测系统、罗氏Cobas e 411及雅培i1000化学发光免疫分析系统，对CEA、ProGRP进行同步检测；评估各系统精密度(批内及总变异系数CV)，分析海尔施系统与雅培、罗氏系统定量结果的相关性(回归方程y = ax + b、相关系数r)及定性判读一致性；采用Bland-Altman (BA)法评价三组系统结果一致性；通过ROC曲线对比各系统单项及联合检测对消化道肿瘤的诊断效能，分析三者临床诊断性能的一致性。结果：三台全自动免疫分析仪各项目的变异系数CV均 ≤ 5%；对于CEA项目，海尔施 vs. 罗氏y = 0.746x + 0.7329，r= 0.818，海尔施 vs. 雅培y = 0.5139x + 1.7938，r = 0.3175，雅培 vs. 罗氏y = 0.6075x + 1.4211，r = 0.4512；对于ProGRP项目，海尔施 vs. 罗氏y = 1.0046x + 4.5166, r = 0.6965，海尔施 vs. 雅培y = 2.2329x + 8.5638，r = 0.7326，雅培 vs. 罗氏y = 0.3897x + 0.2043，r = 0.7132。ROC曲线评价结果显示，海尔施CEA (敏感度32.4%、特异性93.5%、阳性预测值92.3%、阴性预测值36.9%，cutoff值为3.05，AUC为0.600，95% CI为0.537~0.659，P为0.0041)，罗氏CEA (敏感度75.68%、特异性66.23%、阳性预测值84.3%、阴性预测值53.6%，cutoff值为1.42，AUC为0.756，95% CI为0.700~0.807，P < 0.0001)，雅培CEA (敏感度52.43%、特异性87.01%、阳性预测值90.6%、阴性预测值43.6%，cutoff值为2.21，AUC为0.723，95% CI为0.665~0.777，P < 0.0001)。海尔施ProGRP (敏感度68.38%、特异性58.44%、阳性预测值79.7%、阴性预测值44.2%，cutoff值为31.32，AUC为0.663，95% CI为0.591~0.729，P < 0.0001)，罗氏ProGRP (敏感度为83.76%、特异性为70.13%、阳性预测值87.1%、阴性预测值64.7%，cutoff值为27.06，AUC为0.839，95% CI为0.780~0.888，P < 0.0001)，雅培ProGRP (敏感度77.78%、特异性58.44%、阳性预测值81.7%、阴性预测值52.9%，cutoff值为9.01，AUC为0.725，95% CI为0.657~0.787，P < 0.0001)。采用Bland-Altman法对三种仪器检测结果的一致性进行分析，CEA，海尔施 vs. 罗氏，95%的一致性界限的点为97.7%，海尔施 vs. 雅培，95%的一致性界限的点为98.5%，雅培 vs. 罗氏，95%的一致性界限的点为98.1%；ProGRP，海尔施 vs. 罗氏，95%的一致性界限的点为94.7%，海尔施 vs. 雅培，95%的一致性界限的点为96.9%，雅培 vs. 罗氏，95%的一致性界限的点为98.1%。结论：自主研发海尔施CEA、ProGRP检测系统精密度与临床诊断效能接近国际主流同类产品，已具备替代进口设备的基础性能，未来可逐步实现关键技术国产化突破，为临床提供高性价比、稳定可靠的肿瘤标志物检测方案。

Abstract: Objective: To systematically evaluate the methodological performance of a domestically independently developed detection system for carcinoembryonic antigen (CEA) and progastrin-releasing peptide (ProGRP), as well as its efficacy in the early diagnosis of gastrointestinal tumors, and to conduct a parallel comparison with international brands (Abbott and Roche), so as to clarify the real-world application value of this detection system. Methods: In accordance with CLSI EP9-A2 and EP12-A2 guidelines, parallel detection of CEA and ProGRP was performed using the independently developed detection system (Ningbo Health Test Intelligent Manufacturing), the Roche Cobas e 411 chemiluminescent immunoassay system, and the Abbott i1000 chemiluminescent immunoassay system. Precision (intra-assay and total coefficient of variation, CV) of the detection system was determined; quantitative correlation analysis (regression equation y = ax + b and correlation coefficient r) was performed between quantitative results of the in-house system and those of the commercially available Abbott and Roche systems, along with qualitative consistency after positive/negative classification. Bland-Altman (BA) analysis was used to assess the agreement of results among the three detection systems. Finally, receiver operating characteristic (ROC) curves were applied to analyze the individual and combined diagnostic efficacy of the in-house system and imported systems for gastrointestinal tumors, and the consistency of clinical diagnostic performance among the three systems was evaluated. Results: The CV values of all analytes on the three fully automated immunoassay analyzers were ≤ 5%. For CEA: Health Test vs. Roche, y = 0.746x + 0.7329, r = 0.818; Health Test vs. Abbott, y = 0.5139x + 1.7938, r = 0.3175; Abbott vs. Roche, y = 0.6075x + 1.4211, r = 0.4512. For ProGRP: Health Test vs. Roche, y = 1.0046x + 4.5166, r = 0.6965; Health Test vs. Abbott, y = 2.2329x + 8.5638, r = 0.7326; Abbott vs. Roche, y = 0.3897x + 0.2043, r = 0.7132. ROC curve analysis showed the following: Health Test CEA: sensitivity 32.4%, specificity 93.5%, positive predictive value (PPV) 92.3%, negative predictive value (NPV) 36.9%, cutoff value 3.05, AUC 0.600, 95% CI 0.537~0.659, P = 0.0041; Roche CEA: sensitivity 75.68%, specificity 66.23%, PPV 84.3%, NPV 53.6%, cutoff value 1.42, AUC 0.756, 95% CI 0.700~0.807, P < 0.0001; Abbott CEA: sensitivity 52.43%, specificity 87.01%, PPV 90.6%, NPV 43.6%, cutoff value 2.21, AUC 0.723, 95% CI 0.665~0.777, P < 0.0001; Health Test ProGRP: sensitivity 68.38%, specificity 58.44%, PPV 79.7%, NPV 44.2%, cutoff value 31.32, AUC 0.663, 95% CI 0.591~0.729, P < 0.0001; Roche ProGRP: sensitivity 83.76%, specificity 70.13%, PPV 87.1%, NPV 64.7%, cutoff value 27.06, AUC 0.839, 95% CI 0.780~0.888, P < 0.0001; Abbott ProGRP: sensitivity 77.78%, specificity 58.44%, PPV 81.7%, NPV 52.9%, cutoff value 9.01, AUC 0.725, 95% CI 0.657~0.787, P < 0.0001. Bland-Altman analysis for inter-instrument agreement: For CEA, the percentage of points within the 95% limits of agreement was 97.7% (Health Test vs. Roche), 98.5% (Health Test vs. Abbott), and 98.1% (Abbott vs. Roche). For ProGRP, the percentage of points within the 95% limits of agreement was 94.7% (Health Test vs. Roche), 96.9% (Health Test vs. Abbott), and 98.1% (Abbott vs. Roche). Conclusion: The independently developed Health Test CEA and ProGRP detection system exhibits precision and clinical diagnostic performance comparable to mainstream international counterparts, and possesses basic performance for replacing imported products. It can gradually achieve domestic breakthroughs in bottleneck technologies in the future, providing a cost-effective and reliable tumor marker detection solution for clinical practice.