BACKGROUND: BCG vaccination, originally used to prevent tuberculosis, is known to "train" the immune system to improve defence against viral respiratory infections. We investigated whether a previous BCG vaccination is associated with less severe clinical progression of COVID-19 METHODS: A case-control study comparing the proportion with a BCG vaccine scar (indicating previous vaccination) in cases and controls presenting with COVID-19 to health units in Brazil. Cases were subjects with severe COVID-19 (O2 saturation < 90%, severe respiratory effort, severe pneumonia, severe acute respiratory syndrome, sepsis, and septic shock). Controls had COVID-19 not meeting the definition of "severe" above. Unconditional regression was used to estimate vaccine protection against clinical progression to severe disease, with strict control for age, comorbidity, sex, educational level, race/colour, and municipality. Internal matching and conditional regression were used for sensitivity analysis. RESULTS: BCG was associated with high protection against COVID-19 clinical progression, over 87% (95% CI 74-93%) in subjects aged 60 or less and 35% (95% CI - 44-71%) in older subjects. CONCLUSIONS: This protection may be relevant for public health in settings where COVID-19 vaccine coverage is still low and may have implications for research to identify vaccine candidates for COVID-19 that are broadly protective against mortality from future variants. Further research into the immunomodulatory effects of BCG may inform COVID-19 therapeutic research.