In response to the Covid-19 pandemic, governments mobilized scientists to help them design, implement and evaluate policies. These scientists analysed the available scientific evidence and worked to incorporate it into policies, being in this way an essential interface between governments and scientific community. They played the role of "knowledge brokers", filtering, synthesizing or distilling scientific evidence from a broader universe of scientific knowledge - and transforming it into knowledge that could be directly used by governments. In Brazil, in a context of decentralized and fragmented policies against Covid-19, state governments created 69 scientific committees, groups of experts, or crisis committees. This work compares these institutional arrangements, assessing whether - and to what extent - they had the following characteristics: i) timely creation; ii) comprehensive and precise scope; iii) participation of experts; iv) interdisciplinarity; v) transparency; and vi) proximity to the political process. This work also investigates the use of information not based on scientific evidence in government policies or recommendations. These could be related to so-called "early treatment" of Covid-19, other forms of treatment or prevention of Covid-19 not based on scientific evidence, discouragement of social distancing, or support to herd immunity via disease transmission. I compared the characteristics of these institutional arrangements quantitatively, based on which I calculated an indicator that measures the extent to which Brazilian states adopted mechanisms that favoured the incorporation of scientific evidence into policies. Data indicate that most states created good mechanisms of knowledge brokerage. The states with the highest values in the indicator were Paraíba, Rio Grande do Sul, Santa Catarina, São Paulo, Espírito Santo, and Rio Grande do Norte. Other states, however, lacked good mechanisms, such as Alagoas, Amapá, Mato Grosso, and Roraima. Rio de Janeiro showed the greatest instability, having initially a good arrangement, but replacing it by other with limited capabilities. Also, institutional arrangements were better in some points than in others. On the one hand, most governments created these arrangements early in the pandemic, experts played a central role, and arrangements were usually influential. On the other hand, in many cases their scope was not comprehensive, they lacked precision, transparency was limited, and interdisciplinarity was weak. Interdisciplinarity was the weakest point (among those assessed) in state governments' institutional arrangements. Finally, there were in fifteen states policies or recommendations not based on scientific evidence, especially those related to the so-called "early treatment".
