中文摘要：

目的 比较3种以细菌培养为基础的药物敏感性试验（简称“药敏试验”）方法检测结核分枝杆菌对乙胺丁醇的药物敏感性。方法 采用改良罗氏培养基比例法（L-J法）、 Bactec MGIT 960检测系统法（960法）和微孔板Alamar blue显色法［最低抑菌浓度法（MIC法）］同步对126株结核分枝杆菌临床分离株进行乙胺丁醇药敏试验，比较分析3种方法的药敏试验结果。试验数据采用SPSS 17.0软件进行分析处理。Kappa值在0.4～0.75为一致性较好，Kappa值≥0.75为一致性极佳，Kappa值＜0.4为一致性差。结果 3种方法总体一致率为75.4%（95/126）。若以L-J法测定结果为判断标准，则960法和MIC法的敏感度、特异度和一致率分别为62.8%（49/78）、100.0%（48/48）、77.0%（97/126）和82.1%（64/78）、97.9%（47/48）、88.1%（111/126），MIC法与L-J法一致性极佳（Kappa=0.76），而960法与L-J法比较显示一致性较好（Kappa=0.56）。若以960法测定结果为判断标准，则L-J法和MIC法的敏感度、特异度和一致率分别为100.0%（49/49）、62.3%（48/77）、77.0%（97/126）和98.0%（48/49）、77.9%（60/77）、85.7%（108/126），L-J法和MIC法均显示与960法有较好一致性（Kappa≥0.70）。若以MIC法测定耐药结果为判断标准，则L-J法和960法的敏感度、特异度和一致率分别为98.5%（64/65）、77.0%（47/61）、88.1%（111/126）和73.8%（48/65）、98.4%（60/61）、85.7%（108/126）。3种方法的不一致性主要表现在MIC值4.0～16.0μg/ml的药物浓度范围。结论 3种方法对乙胺丁醇进行药敏试验具有较好的一致性，但也存在一定差异。L-J法和960法均与MIC法具有较高一致性，且MIC法性价比更高，适于耐乙胺丁醇结核病的早期诊断。

英文摘要：

Objective To compare the detection effects and the accordance of three drug sensitivity test （DST） methods based on bacterial culture for testing Mycobacterium tuberculosis （M.tuberculosis） sensitivity to ethambutol （EMB）. Methods One hundred and twenty-six M.tuberculosis clinical isolates sensitivity to EMB were detected with three methods, including L-J proportion method, Bactec MGIT 960 system （960 system） and microplate Alamar blue assay （MIC assay）, and their results were compared interactively. The test data were analyzed using SPSS 17.0 software. The Kappa value between 0.4-0.75 represents moderately consistency, and Kappa value ≥0.75 shows excellent concordance, whereas Kappa value less than 0.4 means poor consistency. Results The EMB sensitivity of 126 clinical M.tuberculosis isolates was tested by the three methods, the total consistency was 75.4%（95/126）. If L-J proportion method was used as the standard control, the sensitivity, specificity and consistency of 960 system and MIC assay were 62.8% （49/78）, 100.0% （48/48）, 77.0% （97/126） and 82.1% （64/78）, 97.9% （47/48）, 88.1% （111/126）, respectively. MIC method showed a better consistency with L-J proportion method （Kappa=0.76）, while 960 system represented a moderate concordance with L-J method （Kappa=0.56）. If 960 system were taken as the standard control, the sensitivity, specificity and consistency of 960 system and MIC assay were 100.0% （49/49）, 62.3% （48/77）, 77.0% （97/126） and 98.0% （48/49）, 77.9% （60/77）, 85.7% （108/126）, respectively. Both L-J proportion method and MIC assay showed high consistency with 960 system （Kappa≥0.70）. If MIC assay was set as the standard control, the sensitivity, specificity and consistency of L-J proportion method and 960 system were 98.5% （64/65）, 77.0% （47/61）, 88.1% （111/126）, and 73.8% （48/65）,98.4% （60/61）, 85.7% （108/126）, respectively. The inconsistency among the three methods exis-ted in the MIC v