中文摘要：

目的探讨Xpert MTB/RIF检测痰标本中结核分枝杆菌（MTB）的最佳实验条件并进行方法学评价。方法应用Xpert MTB/RIF检测MBT标准株H37Rv、H37Ra和牛分枝杆菌及20种常见非结核分枝杆菌（NTM）、18种非分枝杆菌病原体和临床分离株,分析Xpert MTB/RIF的敏感性、特异性和重复性,比较不同前处理方法对实验结果的影响,针对仪器出现的报警、错误及无效结果进行相应的处置和总结。结果 MTB标准菌株H37Rv、H37Ra和牛分枝杆菌Xpert MTB/RIF检测阳性,敏感度为1×10^2 CFU/ml（在痰标本中的检测极限为106 CFU/ml）;20种常见NTM和18种非分枝杆菌病原体检测阴性;标本中含有1%浓度的红细胞、白细胞、血清未对103/ml的H37Rv产生干扰作用;重复试验表明本法批间、批内变异系数均小于15%。100株MTB（其中50株Hain线性探针试验确认存在rpoB基因突变）和20株NTM临床分离株加入阴性临床痰标本制备成为含菌浓度为2×102条/ml的模拟痰标本经Xpert MTB/RIF检测均相符。在SR处理液加入半胱氨酸的改良直接法前处理痰液Xpert MTB/RIF检测出现＂错误＋无效＂的比例显著低于沉淀处理法和直接法前处理的痰标本（χ2值分别为4.19和6.74,P〈0.01）。结论 Xpert MTB/RIF灵敏、特异、快速、简便,可用于MTB和利福平耐药MTB的快速检测,但仍存在错误和无效检测而影响实验结果,应采取有效措施予以避免。

英文摘要：

Objective To conduct a methodological study to determine the optimal experimental conditions for use of Xpert MTB/RIF to detect Mycobacterium tuberculosis in sputum specimens. Methods The H37Rv and H37Ra refer- ence strains of M. tuberculosis, strains of M. boris, 20 types of common non-tuberculous mycobacteria （NTM）, 18 types of non-mycobacterial pathogens, and clinical isolates were used along with negative sputum specimens. The sensitivity, specificity and repeatability of Xpert MTB/RIF were analyzed. An interference test was conducted by adding 1% red blood cells, white blood cells, and serum to sputum specimens containing bacteria. The effects of different pre-treatments on the experimental results were compared, and alarms, errors, and invalid results yielded by equipment were dealt with and summarized accordingly. Results Xpert MTB/RIF detected the H37Rv and H37Ra reference strains of M. tuber culosis and strains of M. boris with a sensitivity of 1 ×10^2 CFU/ml （the detection limit for sputum specimens was 106 CFU/ml）. Xpert MTB/RIF did not detect 20 types of common NTM or 18 types of non-mycobaeterial pathogens. Speci- mens containing 1% red blood cells, white blood cells, or serum did not interfere with H37Rv production of 10^3/ml. A re-test indicated that the intra-bateh and inter-batch coefficients of variation were both less than 15%. One hundred strains of M. tuberculosis （50 of those strains had a mutation in the rpoB gene that was verified with a line probe assay） and 20 NTM clinical isolates were added to negative sputum samples to prepare simulated sputum specimens with a con- centration of 2 ×10^2/ml, and Xpert MTB/RIF results accorded with that value. Three different pretreatments were compared. When modified direct pre-treatment of sputum was performed and cysteine was added to sample reagent, Xpert MTB/RIF yielded a markedly lower ratio＂errors＋ invalid results＂ than when precipitation treatment or direct pre-treat-ment of specimens was performed （X2 values were 4. 19 and