Abstract
BACKGROUND: Bedaquiline-containing regimens have been widely used to treat patients with drug-resistant tuberculosis in South Africa since 2019. We aimed to estimate the prevalence of bedaquiline resistance among patients in the Western Cape with rifampicin-resistant tuberculosis tested by Xpert MTB/RIF Ultra (Xpert). METHODS: This prospective study analysed consecutive Mycobacterium tuberculosis diagnostic isolates collected from patients with Xpert-tested rifampicin-resistant tuberculosis in the Western Cape, South Africa, between March 30, 2023, and Jan 3, 2024. We used the Deeplex Myc-TB assay within routine clinical workflows to test genotypic resistance to bedaquiline and other antituberculosis drugs; mmpR5 variants were classified according to Deeplex version 3.0.1 extended catalogue. Phenotypic drug susceptibility information was derived from the National Health Laboratory System and Stellenbosch University for isolates with a Deeplex-identified mmpR5 variant. We estimated the prevalence of bedaquiline resistance and the diagnostic accuracy of Deeplex for bedaquiline susceptibility using a composite genotypic-phenotypic reference standard. FINDINGS: Of 701 sputum sediments, 131 (19%) were culture-negative. Of the 570 isolates accumulated during the study period, Deeplex was not performed for 139 during intervals trialling workflow optimisation procedures. Of 431 isolates, we successfully sequenced 401 (93%), of which 15 (4%) were found to be rifampicin-susceptible; 364 (91%) analysed isolates were baseline and 37 (9%) were longitudinal (median estimated time since previous diagnosis of 5·4 months [IQR 3·7-8·0]). Bedaquiline resistance was detected in 45 (12% [95% CI 9-16]) of 364 baseline and 15 (41% [25-58]) of 37 longitudinal isolates. Deeplex-tested resistance-associated and uncharacterised mmpR5 variants had a similar likelihood of being phenotypic drug susceptibility testing-resistant (37 [97%] of 38 and 16 [94%] of 17, respectively; p=0·53). Combining both types of variants, Deeplex had a sensitivity of 93% (95% CI 83-98) and specificity of 99% (97-100). INTERPRETATION: In a prospective, representative sample of patients with Xpert-tested rifampicin-resistant tuberculosis, we found an elevated prevalence of bedaquiline resistance, particularly in patients with recent tuberculosis treatment. Efficient and accurate surveillance for bedaquiline resistance should be a high programmatic priority. FUNDING: The National Institute of Allergy and Infectious Diseases (at the National Institutes of Health) and Unitaid.