Programmatic diagnostic accuracy and clinical utility of Xpert MTB/XDR in patients with rifampicin-resistant tuberculosis in Georgia

Background: Xpert MTB/XDR (Cepheid, USA) is recommended for drug susceptiblity testing in patients with tuberculosis (TB) by the World Health Organization (WHO) with potential for rapid detection of isoniazid and fluoroquinolones resistance. However, diagnostic accuracy and clinical utility in a programmatic setting are unknown. Methods: We evaluated accuracy and clinical utility of Xpert XDR in patients with rifampicin-resistant pulmonary TB during programmatic implementation in Georgia between July 2022 and August 2024, using phenotypic drug susceptibility testing (pDST) as reference standard. Results: 140 patients were tested with Xpert MTB/XDR and pDST, and 94.9% and 33.8% had isoniazid and fluoroquinolone resistance by pDST respectively. Xpert MTB/XDR showed 99.2% sensitivity (95% CI 95.5-100%) and 100% specificity (95% CI 54.1-100%) for isoniazid resistance. Sensitivity and specificity for fluoroquinolone resistance were 88.4% (95% CI 74.9-96.1%) and 100% (95% CI 95.6-100%). Including indeterminate/invalid Xpert MTB/XDR results, 17.4% (8/46) and 6.4% (8/129) of patients with phenotypic fluoroquinolone and isoniazid resistance were missed, respectively. Median turn-around time for Xpert MTB/XDR was 1 day (IQR 1-3) and median time to treatment was 4 days (IQR 1-7). Phenotypic DST results took a median of 43 days (IQR 29-63) longer than Xpert MTB/XDR results. 95% (115/121, 95% CI 89.5-98.2%) of patients had fluoroquinolones appropriately prescribed based on Xpert XDR result. Conclusions: Progammatic data confirms high accuracy of Xpert MTB/XDR, although below WHO TPP targets for fluoroquinolones, with significantly faster time-to-results than pDST. However, there is a strong need for rapid tests to detect resistance to bedaquiline and other newer TB drugs.