SETTING: Studies have shown that linezolid (LZD) can be used to treat extensively drug-resistant tuberculosis (XDR-TB).OBJECTIVE: To conduct a systematic review and meta-analysis to assess existing evidence concerning efficacy and safety of LZD for XDR-TB treatment.DESIGN: The MEDLINE@OVID, PubMed, EMBASE, the Cochrane Library, Clinical Trials, Sinomed, CMCI, CNKI, VIP and Wanfang databases were systematically searched for randomised controlled trials, cohort studies, case series or case reports on XDR-TB patients treated with LZD from January 2000 to December 2016. Summary estimates of the rate of sputum culture conversion, treatment success and adverse effects were calculated; data that could not be combined were summarised and described qualitatively. The combined results were examined for heterogeneity, sensitivity and publishing bias.RESULTS: Twenty-two original studies covering a total of 302 patients with XDR-TB fulfilled the inclusion criteria. Pooled estimates for sputum culture conversion and treatment success rates were respectively 93.2% and 67.4% in XDR-TB patients on LZD treatment. The pooled estimates for the rate of myelosuppression, peripheral neuropathy, optic neuritis and adverse reactions of the gastrointestinal tract were respectively 42.5%, 26.0%, 19.0% and 35.0%. Heterogeneity was mostly due to the initial dose of LZD (≤600 mg/d or >600 mg/d), as patients with a high initial dose of LZD were more likely to have myelosuppression (48.4% vs. 24.8%, P = 0.010) and adverse events of the gastrointestinal tract (41.3% vs.15.4%, P = 0.100).CONCLUSION: LZD appears to be effective for XDR-TB, but adverse events are common. An LZD dose of ≤600 mg/d as the initial dose for treating XDR-TB patients is recommended.