Abstract Background The WHO-approved Xpert MTB/XDR test detects Mycobacterium tuberculosis and resistance to isoniazid, fluoroquinolones, ethionamide and injectable-drugs directly on specimens. This pragmatic, laboratory-based study assessed diagnostic accuracy and feasibility of a reflex testing approach, where Xpert MTB/XDR was performed on residual specimens previously processed for Xpert MTB/RIF Ultra. Methods Routine respiratory specimens, processed for Xpert MTB/RIF Ultra, were stored in sample reagent buffer at 2-8°C. If rifampicin-resistant, the residual specimen was assessed for volume adequacy (≥2ml) and Xpert MTB/XDR tested, with storage time recorded. A second specimen was used for routine and reference standard testing (culture and sequencing). Results Specimens (99% sputum) from 763 participants submitted to 2 large routine laboratories were included. Xpert MTB/XDR yielded valid resistance detection results in 639 (84%) compared with 507 (66%) for routine testing (difference [95%CI]=18 [13-22]). Median turnaround time for results was 23 hours for Xpert MTB/XDR and 15 days for routine testing. While 748 (98%) specimens were ≥2 ml, only 102 (13%) were stored for ≤4 hours. By reference standard, 284/394 (72%) were isoniazid-resistant and 57/380 (15%) were fluroquinolone-resistant. Xpert MTB/XDR sensitivity was 94% (95%CI 91-97) for isoniazid and 91% (81-97) for fluoroquinolone resistance detection. Specificity was 98% (94-100) and 100% (98-100), respectively. Conclusion Xpert MTB/XDR performed favorably compared to the reference and the reflex testing approach increased results availability over routine testing, while dramatically decreasing turnaround time from weeks to hours. Laboratory workflow precluded testing within the manufacturer-recommended 4-hour storage time, but longer storage did not appear detrimental.