Background: Data on the safety and efficacy profile of tocilizumab in patients with severe COVID-19 needs to be enriched.Methods: In this open label, prospective study, we evaluated clinical outcomes in consecutive patients with COVID-19 and PaO2/FiO2<200 receiving tocilizumab plus usual care versus usual care alone. The primary outcome was 28-day mortality. Secondary outcomes included time to discharge, change in PaO2/FiO2 at day 5 and change in WHO progression scale at day 10.Findings: Overall, 114 patients were included in the analysis (tocilizumab plus usual care: 56, usual care: 58). Allocation to usual care was associated with significant increase in 28-day mortality compared to tocilizumab plus usual care [Cox proportional-hazards model: HR: 3.34, (95%CI: 1.21 to 9.30), (p=0.02)]. There was not a statistically significant difference with regards to hospital discharge over the 28-day period for patients receiving tocilizumab compared to usual care [11.0 days (95%CI: 9.0 to 16.0) vs 14.0 days (95%CI: 10.0 to 24.0), HR: 1.32 (95%CI: 0.84 to 2.08), p=0.21]. ΔPaO2/FiO2 at day 5 was significantly higher in the tocilizumab group compared to the usual care group [42.0 (95%CI: 23.0 to 84.7) vs 15.8 (95%CI: -19.4 to 50.3), p=0.03]. ΔWHO scale at day 10 was significantly lower in the tocilizumab group compared to the usual care group (-0.5±2.1 vs 0.6±2.6, p=0.005). Conclusion: This is the first study administrating tocilizumab in patients with COVID-19 based on PaO2/FiO2. Tocilizumab improved survival and other clinical outcomes in hospitalized patients with COVID-19 and PaO2/FiO2<200 irrespective of systemic inflammatory markers levels.