Rationale: Until 2020, extensively drug-resistant tuberculosis (XDR-TB) was defined as resistance to rifampicin and isoniazid (multidrug-resistant tuberculosis, MDR-TB), any fluoroquinolone (FQ) and any second-line injectable drug (SLID). In 2019 the World Health Organization issued new recommendations for managing patients with drug-resistant tuberculosis, substantially limiting the role of SLID in MDR-TB treatment and thus putting that XDR-TB definition into question. Objective: To propose an up-to-date definition for XDR-TB. Methods: We used a large dataset to assess treatment outcomes for MDR-TB patients exposed to any type of longer regimen. We included patients with bacteriologically confirmed MDR-TB and known FQ and SLID resistance results. We did logistic regression to estimate adjusted odds ratios (aORs) for unfavourable treatment outcome (failure, relapse, death, loss-to-follow-up) by resistance pattern (FQ, SLID) and Group A drug use (moxifloxacin/levofloxacin, linezolid, bedaquiline). Measurements and Main Results: We included 11,666 patients with MDR-TB; 4653 (39.9%) had an unfavourable treatment outcome. Resistance to FQs increased the odds of an unfavourable treatment outcome (aOR 1.91; 95% confidence interval [95%CI] 1.63-2.23). Administration of bedaquiline and/or linezolid improved treatment outcomes regardless of resistance to FQ and/or SLID. Among XDR-TB patients, compared to persons receiving no Group A drug, aORs for unfavourable outcome were 0.37 (95%CI 0.20-0.69) with linezolid only, 0.40 (95%CI 0.21-0.77) with bedaquiline only, and 0.21 (95%CI 0.12-0.38) with both. Conclusions: Our study supports a new definition of XDR-TB as MDR plus additional resistance to FQ plus bedaquiline and/or linezolid, and helps assess the adequacy of this definition for surveillance and treatment choice.