Paucibacillary Tuberculosis Drives the Low Positive Predictive Value of Xpert MTB/RIF Ultra for Rifampicin Resistance Detection in Low-Prevalence Settings.

Summary

In Rwanda, only 32% of Xpert Ultra RR-TB calls were confirmed on repeat testing; among samples with very low bacillary load, 89% of RR calls were false, yielding substantial overtreatment. Programs should confirm RR-TB detected at very low bacillary load (e.g., via repeat Ultra, rpoB sequencing, or pDST) and adjust algorithms accordingly.

Key Findings

• Only 32% (41/129) of initial Ultra RR-TB calls were concordant on repeat Ultra testing.
• Among 'very low' bacillary load samples, 89% had false RR results (risk ratio 8.20; 95% CI 3.56–18.85).
• Overall, 53% (54/101) of patients with reference testing received unnecessary RR-TB treatment due to false resistance calls.

Clinical Implications

When Ultra reports rifampicin resistance with very low bacillary load, require confirmatory testing before initiating RR-TB regimens. Implement repeat Ultra and, where feasible, rpoB sequencing and pDST to improve PPV.

Limitations

• Reference testing was not available for all unconfirmed cases, introducing potential verification bias
• Single-country setting may limit generalizability to other epidemiologic contexts

Future Directions

Develop and validate diagnostic algorithms that incorporate bacillary load thresholds to trigger confirmatory testing; assess cost-effectiveness and patient outcomes of revised workflows.