摘要目的:比较3种血脂检测系统结果的差异，分析其对血脂管理中危险分层及临床决策的影响，并寻找减小影响的方法。方法:在2022年8—10月中南大学湘雅二医院体检者及住院患者中收集甘油三酯（TG）<4.5 mmol/L的血清样本196份，分别用日立-和光（HW）、罗氏、迈瑞3种系统直接检测TG、总胆固醇（TC）、低密度脂蛋白胆固醇（LDL-C）、高密度脂蛋白胆固醇（HDL-C），并由公式（TC-HDL-C）计算非高密度脂蛋白胆固醇（非HDL-C）以及根据Friedewald公式计算LDL-C（F-LDL-C），进行方法学对比。分别计算这6个指标在3种检测系统间的变异系数（ CV），评价结果的一致性，用Pearson相关系数评价各指标在不同系统间的相关性。根据《中国血脂管理指南》推荐的不同危险分层LDL-C的达标值将样本分成<1.4、1.4~<1.8、1.8~<2.6、2.6~<3.4和≥3.4 mmol/L组，统计不同系统LDL-C检测结果分在同一组的样本数及百分率，评估系统间LDL-C差异对血脂管理临床决策的影响。通过2种方法计算校正因子：（1）用EP9-A3的方法估算系统间LDL-C的平均偏差；（2）采用多元线性逐步回归建立系统间LDL-C差值与相关指标的回归模型。用这2种校正因子修正系统间LDL-C的偏差，用卡方检验比较修正前后LDL-C分组一致率的差异。 结果:3种检测系统间TG、TC、LDL-C、F-LDL-C、HDL-C、非HDL-C的 CV均值分别为4.84%、1.92%、11.96%、3.81%、5.82%、2.61%。相关性分析显示，3种系统进行两两比对时，除HW与罗氏的LDL-C、迈瑞与罗氏的LDL-C的 R2分别为0.938、0.947外，其余指标 R2均>0.97。3种系统LDL-C、F-LDL-C按危险分层达标值分组的一致率分别为51.0%（100/196）、90.8%（178/196），差异有统计学意义（ P<0.05）。两两比较时，罗氏与HW、迈瑞与HW、迈瑞与罗氏系统LDL-C分组一致率分别为60.7%（119/196）、82.7%（162/196）、54.1%（106/196）。用平均偏差校正后罗氏与HW的分组一致率提高至73.7%~79.4%（ P<0.05），罗氏与迈瑞的分组一致率升高至72.3%~79.0%（ P<0.05）；用差值回归模型校正后罗氏与HW的分组一致率提高至82.5%~84.0%，罗氏与迈瑞的分组一致率升高至81.0%~89.2%，而用2种校正方法校正后迈瑞与HW的分组一致率无明显变化（ P>0.05）。 结论:不同检测系统之间LDL-C存在明显差异，按降脂达标值分组的一致率较低，可能影响血脂管理的临床决策。通过校正因子修正后可提高LDL-C差异较大的罗氏与HW、罗氏与迈瑞系统间分组的一致率。用差值多元线性回归模型作为校正因子优于平均偏差。

abstractsObjective:The results of the three lipid detection systems were compared to analyze their influence on risk stratification and clinical treatment in lipid management, especially the target goal cut-off point determination, and to find ways to reduce the impact on target goal determination of various lipid measurement system.Methods:A total of 196 serum samples with triglyceride TG <4.5 mmol/L were collected from people undergoing physical examinations and in-patients in the Second Xiangya Hospital of Central South University from August to October 2022. Triglyceride (TG), total cholesterol (TC), low-density lipoprotein cholesterol (LDL-C) and high-density lipoprotein cholesterol (HDL-C) were directly detected with Hitachi-Woke (HW), Roche and Mindray detection systems, respectively. The non high-density lipoprotein cholesterol (non HDL-C) was calculated by formula (TC-HDL-C) and LDL-C (F-LDL-C) was calculated by Friedewald formula, and results from various methodology were compared. The coefficient of variation ( CV) of these six indicators derived from the three detection systems were calculated to evaluate the consistency of the obtained results from different venders. In addition, the Pearson correlation coefficient was analyzed to evaluate the correlation of each indicator among different systems. According to the Chinese Guidelines for Blood Lipid Management, samples were divided into groups with LDL-C levels of <1.4, 1.4-<1.8, 1.8-<2.6, 2.6-<3.4 and ≥3.4 mmol/L according to the recommended LDL-C levels for different risk stratification levels. The sample size and percentage of LDL-C test results from different systems in the same group were counted to evaluate the impact of LDL-C differences between systems on clinical decision-making of blood lipid management. The correction factor was calculated through two methods: (1) The average deviation of LDL-C between systems was estimated by EP9-A3 method; (2) Multiple linear stepwise regression was used to establish the regression model of LDL-C difference and related indexes between systems. The two correction factors were used to correct the deviation of LDL-C value obtained from various systems, and Chi-square test was used to compare the difference of LDL-C grouping consistency rate before and after correction. Result:The average CV values of TG, TC, LDL-C, F-LDL-C, HDL-C, and non HDL-C among the three detection systems were 4.84%, 1.92%, 11.96%, 3.81%, 5.82% and 2.61%, respectively. Correlation analysis showed that when comparing the three systems in pairs, except for LDL-C derived from HW and Roche′s, and Mindray and Roche′s LDL-C ( R 2=0.938 and 0.947), the R 2 of other indicators were all greater than 0.97. The consistency rates of the three systems on LDL-C and F-LDL-C were 51.0% (100/196) and 90.8% (178/196), respectively, according to the risk stratification standard values and the difference was statistically significant ( P<0.05). When comparing in pairs, the consistency rates of Roche and HW, Mindray and HW, Mindray and Roche system LDL-C grouping were 60.7% (119/196), 82.7% (162/196), and 54.1% (106/196), respectively. After adjusting for mean deviation, the group consistency rate of Roche and HW increased to 73.7%-79.4% ( P<0.05), and the group consistency rate of Roche and Mindray increased to 72.3%-79.0% ( P<0.05). After adjusting for difference regression model, the group consistency rate of Roche and HW increased to 82.5%-84.0%, and the group consistency rate of Roche and Mindray increased to 81.0%-89.2%. However, there was no significant change in the group consistency rate of Mindray and HW after adjusting for both correction methods ( P>0.05) .Conclusions:There are significant differences in LDL-C derived from different detection systems, and the consistency rate of grouping according to the lipid-lowering standard value is relatively low, which may affect clinical decision-making in lipid management. Adjusted by the correction factor, the consistency rate of grouping between Roche and HW, Roche and Mindray systems with large differences in LDL-C can be improved. Using the difference multiple linear regression model as a correction factor is superior to the average deviation.