In 2008-2009 a RVF outbreak occurred in the Anjozorobe area, a temperate and mountainous area of Madagascar highlands. A serological study conducted there in 2009 showed an IgG seroprevalence rate of 28%. Data analysis suggested a recurrent circulation of RVFV. The objectives of this study were to describe the cattle trade network in this area and analyze the link between network structure and RVFV circulation. Questionnaire survey among 386 breeders from 47 villages was carried out to collect trade data. Yearly village-level seroconversion rate was estimated in 2010 by testing 516 cattle negative in 2009. Association between the occurrence of seroconversion and network centrality parameters, distance to the nearest water point and 2009 seroprevalence level was tested. Due to the non-independence of the centrality parameters, a bootstrap procedure was used to assess the effects of the independent variables. Average village-level seroconversion was 7%, ranging from 0-20%. Two types of trading practice were observed: exchanges and buy/sale. The corresponding networks appeared both scale-free, and a significant but low correlation was observed between them. A negative association was observed between the occurrence of seroconversion in the village and the 2009 seroprevalence level, as well as the distance to the nearest water point. After RVFV introduction, vector-based transmission may support the within-village circulation. The node degree in the exchange network was positively linked with the occurrence of seroconversion. It was not the case for the buy/sale network. Both networks could thus have distinct roles in RVFV circulation. The exchanges network could be the support for RVFV introduction in villages, the buy/sale network being probably rather implicated in the introduction of RVFV in the area, from other parts of Madagascar. (Texte intégral)