Abstract
The Roche cobas MTB and MTB-RIF/INH assays allow for detection of Mycobacterium tuberculosis complex (MTBC) nucleic acid and rifampicin (RIF) and isoniazid (INH) resistance-associated mutations in an automated, high-throughput workflow. In this study, we evaluated the performance of these assays, employing samples from settings of low and high tuberculosis (TB) burdens. A total of 325 frozen, leftover respiratory samples collected from treatment-naive patients with presumptive TB in Germany (n = 280) and presumptive RIF-resistant TB in Sierra Leone (n = 45) were used in this study. cobas MTB results for detection of MTBC DNA from N-acetyl-l-cysteine-sodium hydroxide (NALC-NaOH)-treated samples were compared to culture results. Predictions of RIF and INH resistance by the cobas MTB-RIF/INH assay were compared to a composite reference standard (phenotypic drug susceptibility testing and line probe assay). Whole-genome sequencing was used to resolve discordances. The overall sensitivity of cobas MTB for detection of MTBC DNA in culture-positive samples (n = 102) was 89.2% (95% confidence interval [CI], 81.7 to 93.9%). The specificity of cobas MTB was 98.6% (95% CI, 96.1 to 99.5%). Sensitivity and specificity for detection of RIF and INH resistance were 88.4% (95% CI, 75.5 to 94.9%) and 97.6% (95% CI, 87.4 to 99.6%) and 76.6% (95% CI, 62.8 to 86.4%) and 100.0% (95% CI, 90.8 to 100.0%), respectively. Discordant results for RIF and INH resistance were mainly due to uncommon mutations in samples from Sierra Leone that were not covered by the cobas MTB-RIF/INH assay. In conclusion, cobas MTB and MTB-RIF/INH assays provide accurate detection of MTBC DNA and resistance-associated mutations in respiratory samples. The influence of regional variations in the prevalence of resistance-conferring mutations requires further investigation.