Prevalence and drivers of false-positive rifampicin-resistant Xpert MTB/RIF results: a prospective observational study in Rwanda

Abstract

Summary: Background: The Xpert MTB/RIF (Xpert) assay is used globally to rapidly diagnose tuberculosis and resistance to rifampicin. We investigated the frequency and predictors of false-positive findings of rifampicin resistance with Xpert. Methods: We did a prospective, observational study of individuals who were enrolled in a Rwandan nationwide diagnostic cohort study (DIAMA trial; NCT03303963). We included patients identified to have rifampicin resistance on initial Xpert testing. We did a repeat Xpert assay and used rpoB Sanger and deep sequencing alongside phenotypic drug susceptibility testing (pDST) to ascertain final rifampicin susceptibility status, with any (hetero)resistant result overriding. We used multivariable logistic regression to assess predictors of false rifampicin resistance on initial Xpert testing, adjusted for HIV status, tuberculosis treatment history, initial Xpert semi-quantitative bacillary load, and initial Xpert probe. Findings: Between May 4, 2017, and April 30, 2019, 175 people were identified with rifampicin resistance at initial Xpert testing, of whom 154 (88%) underwent repeat Xpert assay. 54 (35%) patients were confirmed as rifampicin resistant on repeat testing and 100 (65%) were not confirmed with resistance. After further testing and sequencing, 121 (79%) of 154 patients had a final confirmed status for rifampicin susceptibility. 57 (47%) of 121 patients were confirmed to have a false rifampicin resistance result and 64 (53%) had true rifampicin resistance. A high pretest probability of rifampicin resistance did not decrease the odds of false rifampicin resistance (adjusted odds ratio [aOR] 6·0, 95% CI 1·0–35·0, for new tuberculosis patients vs patients who needed retreatment). Ten (16%) of the 64 patients with true rifampicin resistance did not have confirmed rifampicin resistance on repeat Xpert testing, of whom four had heteroresistance. Of 63 patients with a very low bacillary load on Xpert testing, 54 (86%) were falsely diagnosed with rifampicin-resistant tuberculosis. Having a very low bacillary load on Xpert testing was strongly associated with false rifampicin resistance at the initial Xpert assay (aOR 63·6, 95% CI 9·9–410·4). Interpretation: The Xpert testing algorithm should include an assessment of bacillary load and retesting in case rifampicin resistance is detected on a paucibacillary sputum sample. Only when rifampicin resistance has been confirmed on repeat testing should multidrug-resistant tuberculosis treatment be started. When rifampicin resistance has not been confirmed on repeat testing, we propose that patients should be given first-line anti-tuberculosis drugs and monitored closely during treatment, including by baseline culture, pDST, and further Xpert testing. Funding: The European & Developing Countries Clinical Trials Partnership 2 programme, and Belgian Directorate General for Development Cooperation.