Accuracy of molecular diagnostic tests for drug-resistant tuberculosis detection in China: a systematic review

OBJECTIVE : To evaluate the accuracy of molecular diagnostics for the detection of drug-resistant tuberculosis (TB) in Chinese patients. METHOD: Seven databases were searched for eligible studies that evaluated the accuracy of molecular diagnostics against drug susceptibility testing (DST) for detecting drug resistance. A bivariate random-effects meta-analysis was conducted to pool sensitivity and specificity by the index test and drug resistance type. RESULTS : A total of 159 studies were included. Compared with DST (reference standard), Xpert (R) could diagnose rifampicin (RMP) resistant TB accurately, with a pooled sensitivity and pooled specificity of 92% (95% CI 90-94) and 98% (95% CI 97-98), respectively. Line-probe assays (LPAs) also performed well for RMP resistance, with a pooled sensitivity of 91% (95% CI 88-93) and pooled specificity of 98% (95% CI 96-99), but not for isoniazid (INH) or second-line drugs due to lower sensitivity (<80%). The pooled sensitivity of GeneChip (R) microarrays for RMP, INH and multidrug resistance was 89% (95% CI 86-91), 79% (95% CI 75-82) and 79% (95% CI 73-84), respectively, and the specificities were all >97%. Similarly, the MeltPro (R) TB/STR assay had better sensitivity and specificity for first-line drugs, varying from 87% to 89% and 97% to 98%, respectively, than for second-line drugs. CONCLUSION: The Xpert assay, LPA, GeneChip assay, and MeltPro assay are credible methods with high accuracy for RMP resistance detection, but they may not be appropriate for other anti-tuberculosis drugs due to low sensitivity.