Setting: Xpert® MTB/RIF is the first-line diagnostic test for Mycobacterium tuberculosis and rifampicin (RIF) resistance in South Africa.

Objective: To describe the rates of Xpert RIF resistance not confirmed on follow-up testing, as well as the patient and test characteristics associated with discordant results.

Design: Retrospective review of patients with isolates showing Xpert RIF resistance. Line-probe assay, phenotypic drug susceptibility testing or repeat Xpert were all considered confirmatory tests of RIF resistance. 'Discordance' was defined as a patient with RIF resistance identified on initial Xpert testing, with a subsequent confirmatory test indicating RIF susceptibility. Associations were analysed using Pearson χ² difference of proportions and modified Poisson regression.

Results: RIF discordance occurred in 22/263 subjects and was associated with Xpert probe B, probe binding delay, as opposed to probe dropout, and probe binding delays (ΔCt) of between 4 and 4.9.

Conclusion: Discordant RIF resistance was common in our cohort and was associated with Xpert probe delay and use of probe B.