五种自动化免疫学方法和液相色谱串联质谱方法测定25羟维生素D的比较

Comparison of five automated 25OHD immunoassays and liquid chromatography tandem mass spectrometry method

目的比较5种自动化免疫学方法测定25羟维生素D（250HD）和液相色谱串联质谱法（LC-MS／MS）测定250HD方法的一致性。方法本研究属于方法学比对。2014年5月至7月选择245份临床血清样本（总250HD浓度范围：2．8ng／ml-64．0ng／ml，其中154份不含250HD2，91份含有不同浓度的250HD2），应用雅培、索林、IDS、罗氏、西门子五个厂家（文中依次以A，B，C，D，E代表）的自动化免疫学方法和本实验室建立的LC-MS／MS方法共同测定，以LC—MS／MS方法作为参照标准，比较测定结果的一致性，并以250HD〈20ng／ml、20—30ng／ml、〉30ng／ml为临床判定维生素D缺乏、不足和充足的标准，判定各种方法的符合率。测定临床样本之前对各方法进行精密度评价。用MedCalc软件统计处理数据，采用Bland—Ahaman图比较2种方法的差异，以Passing＆Bablok回归分析2种方法之间的相关性。结果6种方法测定245例样本的中位数（2．5％-97．5％范围）分别为：23．5（5．8—4J4．2）ng／ml（LC-MS／MS），20．6（7．1～43．5）ng／ml（A），19．0（5．4—38．0）ng／ml（B），23．0（10．0～38．1）ng／ml（C），20．1（5．1～46．0）ng／ml（D），31．3（12．3—71．1）ng／ml（E），Passing＆Bablok回归分析显示B方法与LC-MS／MS相关性最好（r=0．894），而A、C和D方法与LC—MS／MS平均偏差相对较小，E方法与LC-MS／MS平均偏差最大。在不含250HD2的样本中，5种自动化免疫学方法得出的结果与LC-MS／MS方法得出的结果的相关性均在0．84以上，相关性最好的为B（r=0．930），但当样本中含有250HD2时，各种方法与LC—MS／MS的相关性均下降。以统一的临床判定阈值，与LC—MS／MS得到的临床判定结果相比，A、B、C、D和E的一致率分别为68．6％、64．9％、67．8％、70．6％和51．8％。结论不同检测系统检测结果相差较大，250HD2对各种免疫学方法均有较大影响。不同检测系统临床判定一致率均较低，可能需要建立各自相应的判定阈值。

Objective To compare the concordance of five automated 25OHD immunoassays with liquid chromatography tandem mass spectrometry method （LC -MS/MS）. Methods During May to July in 2014, 245 clinical serum samples that requested 25OHD tests were selected, with a total 25OHD range of 2.8 ng/ml - 64.0 ng/ml, in which 154 samples did not contain 25OHD2 and 91 samples contains both 25OHD2 and 25OHD3. To used a LC - MS/MS method that built in our laboratory to measure 25OHD, five commercial automated chemiluminescent immunoassays from Abbott Diagnostics （A）, DiaSorin LIASON （ B）, IDS-iSYS （ C ）, Roche Diagnostics （ D ）, and Siemens ADVIA Centaur （ E ）. Taking the reference method LC-MS/MS as a standard, to compared the concordance and performance of the five automated 25OHD immunoassays. And used the commonly accepted cutoffs for 25OHDdefieieney （ 〈 20 ng/ml） , and insufficiency （20 - 30 ng/ml）, and sufficiency （ ≥ 30 ng/ml ） to compare the uniformity of differentmethods. Statistical analysiswere performed by MedCalc software, Passing ＆ Bablok regression, Bland ＆ Ahaman plots and Box and whisker plots were performed to compare the differences of the methods. Results The medium （range：2.5% -97.5% ） 25OHD of the 245 serum samples of the six methods was 23.5 （5.8-44.2） ng/ml（LC-MS/MS）,20.6 （7.1-43.5）ng/ml（A） ,19.0 （5.4-38.0） ng/mL （B）,23.0 （10.0-38. 1） ng/ml（C）,20. 1 （5. 1-46.0） ng/ml （D）,31.3 （12.3-71. 1） ng/ml （E）, respectively. Passing and Bablok regression showed that method B had the best correlation coefficient with LC-MS/MS （r = 0. 894）, while methods A, C and D had relatively small bias compared withLC-MS/MS and method E had the large bias. If the serum samples did not contain 25OHD2, all the five automated immunoassays correlated well with LC-MS/MS with a correlation coefficient higher than 0.84, and B has the best correlation with LC-MS/MS （ r = 0. 930）. While all the correlation coefficient between immunoassays and LC-MS/MS decreasedwhen analyzing the samplescontaining 25OHD2. Using the clinical cutoffs, A, B, C, D and E had a concordance of 68.6% , 64.9% , 67.8% , 70.6% and 51.8% compared with LC-MS/ MS, respectively. Conclusions There are significant differences between different detection systems of 25OHD. All the immunoassays results were affected by the existence of 250HD2. The concordance of serum 25OHD resultswas poor between different methods, and it may be necessary to built exclusive cutoffs for different methods.