OBJECTIVES: The SF-6D is a preference-based measure of health developed to estimate utility values from the SF-36. The aim of this study was to estimate preference weights for SF-6D health states representing the preferences of a sample of Southern Brazilian general population. METHODS: A sample of 248 health states defined by the SF-6D has been valued by a  sample of Southern Brazilian population using the standard gamble (SG) method. SG responses were used to estimate regression models at the individual and mean levels to predict preference values for all SF-6D health states. The models were compared with those described in the UK study. RESULTS: Five hundred twenty-eight participants were interviewed, but 146 (28%) were excluded due to inconsistent SG responses. Data from 382 subjects were used to estimate the models, rendering 2224 health states valuations. All Brazilian models have a large number of significant coefficients and a mean absolute difference between observed and predicted values below 0.07. Inconsistent coefficients have been merged to produce the final recommended model. Compared to UK data, Brazilian health state values were lower, leading to a lower constant term in the models.  The best model fitted to Brazilian data was a random effects model using only the main effects variables, different from the preferred British SF-6D mean model, highlighting the importance to adopt a country-specific algorithm in predicting SF-6D health states values. CONCLUSIONS: The results provide the first population-based value set for health states in Brazil, making it possible to generate QALYs for cost-utility studies using regional data. Utility scores based on Brazilian preferences values can be derived from existing SF-36 data sets.